Multilaboratory Validation of First Action Method 2016.04 for Determination of Four Arsenic Species in Fruit Juice by High-Performance Liquid Chromatography–Inductively Coupled Plasma–Mass Spectrometry

Abstract

Before being designated AOAC First Action Official Method^SM 2016.04, the U.S. Food and Drug Administration’s method, EAM 4.10 High Performance Liquid Chromatography–Inductively Coupled Plasma–Mass Spectrometric Determination of Four Arsenic Species in Fruit Juice, underwent both a single-laboratory validation and a multilaboratory validation (MLV) study. Three federal and five state regulatory laboratories participated in the MLV study, which is the primary focus of this manuscript. The method was validated for inorganic arsenic (iAs) measured as the sum of the two iAs species arsenite [As(III)] and arsenate [As(V)], dimethylarsinic acid (DMA), and monomethylarsonic acid (MMA) by analyses of 13 juice samples, including three apple juice, three apple juice concentrate, four grape juice, and three pear juice samples. In addition, two water Standard Reference Materials (SRMs) were analyzed. The method LODs and LOQs obtained among the eight laboratories were approximately 0.3 and 2 ng/g, respectively, for each of the analytes and were adequate for the intended purpose of the method. Each laboratory analyzed method blanks, fortified method blanks, reference materials, triplicate portions of each juice sample, and duplicate fortified juice samples (one for each matrix type) at three fortification levels. In general, repeatability and reproducibility of the method was ≤15% RSD for each species present at a concentration >LOQ. The average recovery of fortified analytes for all laboratories ranged from 98 to 104% iAs, DMA, and MMA for all four juice sample matrixes. The average iAs results for SRMs 1640a and 1643e agreed within the range of 96–98% of certified values for total arsenic.

Arsenic is widely distributed in the environment from natural sources such as volcanic activity and erosion and from anthropogenic activities, including the burning of fossil fuels, mining, ore smelting, and the use of arseniccontaining pesticides, herbicides, and wood preservatives (1, 2). Because of its presence in soil and water, small amounts of arsenic are found in foods. Arsenic in the environment and in food exists in many chemical forms; however, from a public health perspective, arsenic may be categorized broadly as inorganic arsenic (iAs), the primary toxic form, or as organic arsenic, which is generally considered less toxic. iAs species, specifically arsenite [As(III)] and arsenate [As(V)], are known carcinogens, and the consumption of iAs has also been associated with noncancer health effects such as cardiovascular disease, neurotoxicity, and diabetes (2). Organic forms, including monomethylarsonic acid (MMA), dimethylarsinic acid (DMA), and arsenosugars, are considered significantly less toxic than iAs, whereas some organic forms commonly found in seafood, such as arsenobetaine and arsenocholine, are considered nontoxic (3).

Recently, the Joint Food and Agriculture Organization/World Health Organization Expert Committee on Food Additives concluded that in addition to drinking water, food can be a significant contributor to iAs exposure (4). Drinking water contains primarily iAs, and the U.S. Environmental Protection Agency set a regulatory limit for total arsenic in drinking water at 10 μg/L (5). For many years, the U.S. Food and Drug Administration (FDA) has been monitoring total arsenic concentrations in a variety of foods, including fruit juices, through the Total Diet Study program (6). Some fruit juices, however, have been shown to contain significant amounts of less-toxic organic arsenic compounds, primarily DMA and MMA, in addition to iAs (7). Therefore, the quantitation of individual arsenic species of varying toxicity in fruit juice samples is necessary to improve estimates of dietary exposure and evaluate health impacts of iAs in fruit juices.

HPLC coupled with inductively coupled plasma (ICP)-MS is the primary method used for arsenic speciation and is well reviewed (8). Anion-exchange chromatography is often used to separate the four main arsenicals, i.e., As(III), As(V), DMA, and MMA, whereas ICP-MS offers ultralow detection limits capable of quantitating each arsenical at the low levels required for this method. Both ion-exchange and ion-pair chromatography have been used with ICP-MS detection for arsenic speciation in juice (9–11), but to our knowledge, these methods have not been subjected to a multilaboratory validation (MLV) study. In 2012, Conklin and Chen at the FDA published an ion exchange method and single-laboratory validation study for arsenic speciation in fruit juice (10). After some minor modifications, this method was added to the FDA Elemental Analysis Manual for Food and Related Products (EAM) as 4.10 High Performance Liquid Chromatography–Inductively Coupled Plasma–Mass Spectrometric Determination of Four Arsenic Species in Fruit Juice, version 1.0 (EAM 4.10; 12) and was used as the basis for this MLV study, as well as subsequent FDA juice survey assignments (13). The data obtained from FDA juice survey assignments were used to generate the health hazard evaluation by the FDA and led to the proposed action level of 10 ng/g iAs in apple juice (14). The goals of this MLV study were to evaluate the accuracy, precision, and robustness of the method; to ensure that it meets the needs of the agency; and to fulfill the requirements for a Level 3–validated method according to the Guidelines for the Validation of Chemical Methods for the FDA Foods and Veterinary Medicine Program (15).

Collaborative Study

Eight laboratories participated in this study. Prior to participating, selected personnel from each laboratory participated in a training exercise. This was necessary because the majority of the analysts had experience with total elemental analysis using ICP-MS, but few had experience using HPLC. Laboratories were instructed to obtain the necessary equipment for the method, obtain juice samples, analyze them for total arsenic, and determine the concentrations of the individual arsenic species using EAM 4.10. Each laboratory was encouraged to coordinate with the method authors to ensure their readiness before beginning the validation exercise. When analysts were comfortable with the method and achieved satisfactory mass balances for their samples, they were able to start the validation exercise.

Four common juice products were chosen for the exercise: single-strength or ready-to-drink (RTD) apple juice, apple juice concentrate, RTD grape juice, and RTD pear juice. Within the apple juice concentrates, two were collected by FDA investigators, whereas the third concentrate was purchased from a local market (designated AJC1, AJC2, and AJC3, respectively). Of the grape juice samples, two were classified as purple (GJ1 and GJ2) and two as white (GJ3 and GJ4). The RTD apple juice samples were designated AJ1, AJ2, and AJ3, whereas the RTD pear juice samples were designated PJ1, PJ2, and PJ3. The type of each juice was known to the participating laboratories. Unless specified, all juice samples were purchased from local Cincinnati, OH, markets; within each juice type, brands were unique. Samples were not screened for total arsenic or arsenic species before fortification by the organizing laboratory. Selected samples were fortified with various combinations of iAs, DMA, and/or MMA, and values were not disclosed to participating laboratories (Table 1, fifth column). Levels were selected to mimic those reported in the literature and those previously found in various FDA-related surveys of products. Samples were mixed thoroughly and kept refrigerated until needed for shipment. Due to the long window of time allowed for participation, samples sent out after more than 6 months from the time of preparation were stored frozen and then thawed by the organizing laboratory before shipping in a refrigerated condition to participating laboratories. All laboratories were sent 50–60 g each of the 13 juice samples. In cases in which cost or availability was an issue, the central laboratory provided additional supplies such as arsenobetaine (AsB), aliquots of Standard Reference Material (SRM) 1643e, mobile phase salt, and PRP-X100 guard columns. During the validation study, procedural clarifications and instrument-specific queries were addressed by the method authors.

Table 1.

Summary of mean concentration, repeatability, and reproducibility data for analysis of RTD apple, grape, and pear juice and apple juice concentrate samples

Analyte	Matrix	No. of Labs	No. of replicates	Analyte added, μg/kg	Mean concn, μg/kg	sr	sR	RSDr, %	RSDR, %	HorRat
iAsa	AJ1	8	24		11.5	0.34	0.81	3.0	7.0	0.22
	AJ2	8	24	25 As(V)	24.8	0.96	2.66	3.9	10.7	0.38
	AJ3	8	24		5.63	0.24	0.59	4.3	10.5	0.30
	GJ1	8	24		5.96	0.25	0.67	4.2	11.3	0.33
	GJ2	8	23		14.4	0.31	0.57	2.1	3.9	0.13
	GJ3	8	24		12.4	0.45	0.60	3.6	4.9	0.16
	GJ4	8	24	10 As(V)	23.2	0.80	1.63	3.4	7.0	0.25
	PJ1	8	24		5.92	0.26	0.68	4.4	11.5	0.33
	PJ2	8	24	10 As(III)	14.7	0.48	0.97	3.2	6.6	0.22
	PJ3	8	24		5.85	0.26	0.70	4.5	12.0	0.35
	AJC1	8	24		27.2	1.40	2.61	5.2	9.6	0.35
	AJC2	8	24		TR (9.94)b	0.43	2.71	4.4	27.2	0.85
	AJC3	8	24		TR (4.94)	0.90	3.26	18.2	66.0	1.86
DMA	AJ1	7	21		<0.25	NAc	NA	NA	NA	NA
	AJ2	7	21		<0.25	NA	NA	NA	NA	NA
	AJ3	7	21		<0.25	NA	NA	NA	NA	NA
	GJ1	7	21		TR (0.94)	0.20	0.31	21.8	33.3	0.73
	GJ2	7	20	30 DMA	28.8	0.95	1.9	3.3	6.7	0.24
	GJ3	7	21		TR (1.12)	0.11	0.31	9.8	27.7	0.62
	GJ4	7	21		TR (0.78)	0.08	0.25	10.9	31.7	0.68
	PJ1	7	21		TR (0.93)	0.11	0.25	12.0	27.5	0.60
	PJ2	7	21		TR (0.48)	0.26	0.55	54.2	115.6	2.3
	PJ3	7	21		<0.25	NA	NA	NA	NA	NA
	AJC1	7	21		TR (4.34)	0.45	2.18	10.4	50.2	1.39
	AJC2	7	21		<1.5	NA	NA	NA	NA	NA
	AJC3	7	21		<1.5	NA	NA	NA	NA	NA
MMA	AJ1	7	21		<0.25	NA	NA	NA	NA	NA
	AJ2	7	21	25 MMA	22.5	0.71	2.09	3.1	9.3	0.33
	AJ3	7	21		TR (0.79)	0.05	0.46	6.5	58.2	1.24
	GJ1	7	21		<0.25	NA	NA	NA	NA	NA
	GJ2	7	21		<0.25	NA	NA	NA	NA	NA
	GJ3	7	21		<0.25	NA	NA	NA	NA	NA
	GJ4	7	21	10 MMA	9.11	0.32	0.69	3.5	7.6	0.23
	PJ1	7	21	30 MMA	27.0	0.61	1.95	2.2	7.2	0.26
	PJ2	7	21		<0.25	NA	NA	NA	NA	NA
	PJ3	7	21		<0.25	NA	NA	NA	NA	NA
	AJC1	7	21		55.5	2.42	3.33	4.4	6.0	0.24
	AJC2	7	21		< 1.5	NA	NA	NA	NA	NA
	AJC3	7	21		< 1.5	NA	NA	NA	NA	NA

^aMeasured as As(III) + As(V).

^bTR = Trace; a result >LOD but ≤LOQ is reported as TR, with the value in parentheses.

^cNA = Not applicable.

Each laboratory was provided with procedural guidance to supplement EAM 4.10, including specific instructions, e.g., regarding the number of replicates to analyze and the approximate fortification levels to use. A reporting template accompanied these instructions to unify data reporting. Upon request, data packets including calculations and raw data printouts were sent to the organizing laboratory. All laboratories were instructed not to discuss their results with other participating laboratories.

The participating laboratories included three FDA laboratories and five state laboratories, for a total of eight data sets including the organizing laboratory. Participating state laboratories were all part of the Food Emergency Response Network (16). Data sets were not edited to ensure the most complete representation; however, two laboratories exhibited a calculation issue that required reprocessing of their initial data (explained later). This manuscript describes the MLV exercise carried out by the eight laboratories using EAM 4.10, which has recently been granted AOAC First Action Status (17), to analyze apple, pear, and grape juices for iAs, DMA, and MMA.

Apparatus and Reagents

Details regarding the apparatus and reagents can be found in AOAC First Action Official Method^SM 2016.04 (17). Deviations from, or clarifications to, that method are explained below. All reagents should be of the highest purity available, e.g., Optima grade (or equivalent), to ensure the lowest blank levels of arsenic possible. National Institute of Standards and Technology (NIST) SRM 1643e Trace Elements in Water and/or SRM 1640a Trace Elements in Natural Water were used to evaluate the accuracy of the method. To our knowledge, a fruit juice SRM for total arsenic and/or arsenic species was not available at the time of this MLV study.

Multiple vendors, depending on the laboratory, were used to supply the various standards and reagents. However, ammonium phosphate dibasic [(NH₄)₂HPO₄] was purchased from Sigma-Aldrich (St. Louis, MO), and due to trace-but-detectable levels of arsenic in other sources, this source was used by all laboratories during the MLV study.

Preparation of Sample and Standard Solutions

Details regarding sample and standard preparation can be found in AOAC First Action Official Method 2016.04 (17). Deviations or clarifications of that method are explained below.

(a). Fortified analytical portions (FAPs) for RTD samples.—

For this validation, duplicates of three levels were prepared for samples AJ1, PJ1, and GJ1 by the following method: Prepare an analytical portion fortified with As(III), DMA, MMA, and As(V) at levels of 5, 25, and 50 μg/kg each by combining 2 g RTD juice and appropriate amounts of the 1000 ng/g multianalyte spiking solution (or a diluted version of this solution) in a 15 mL polypropylene centrifuge tube. Dilute to 10 g total with deionized water (DIW) and mix thoroughly (the spiking levels are 1, 5, and 10 ng/g each in this solution, respectively). Draw an approximately 4 mL aliquot of the analytical solution into a syringe and dispense through a 0.45 μm nylon or PTFE syringe filter (discarding approximately the first 1 mL to waste) into a 15 mL polypropylene centrifuge tube. Fill an autosampler vial approximately half full before analysis. Store the unused portion up to 48 h at 4°C in the event the sample needs to be reanalyzed.

(b). FAPs for juice concentrates.—

For this validation, duplicates of three levels were prepared for sample AJC1 by the following method: Prepare an analytical portion fortified with As(III), DMA, MMA, and As(V) at levels of 35, 100, and 200 μg/kg each by combining approximately 1 g concentrate and appropriate amounts of the 1000 ng/g multianalyte spiking solution in a 15 mL polypropylene centrifuge tube. Dilute to 6 g total with DIW and mix thoroughly. Next, transfer a 2 g portion of this solution to another 15 mL polypropylene centrifuge tube and dilute to 10 g total with DIW and mix thoroughly (the spiking levels are 1.2, 3.3, and 6.7 ng/g each in these solutions, respectively). Draw an approximately 4 mL aliquot of the analytical solution into a syringe and dispense through a 0.45 μm nylon or PTFE syringe filter (discarding the first approximately 1 mL to waste) into a 15 mL polypropylene centrifuge tube. Fill an autosampler vial approximately half full before analysis. Store the unused portion up to 48 h at 4°C in the event the sample needs to be reanalyzed.

(c). Method blank.—

Take 2 g DIW through the sample preparation procedures described above for RTD juices and juice concentrates.

HPLC-ICP-MS Analysis, Quantification, and Identification

Details regarding the analysis of the samples and the quantification and identification of arsenic species can be found in First Action Official Method 2016.04 (17).

Results and Discussion

Data Reporting and Statistical Analysis

Laboratories reported their data by populating the reporting template with the proper data. The central laboratory compiled the data into one spreadsheet and examined it for validity. Upon examining the data, two laboratories had incorrectly calculated the concentrations of their DMA and MMA standards; these laboratories reprocessed their data and resubmitted. Data were omitted from one laboratory for failure to properly determine and verify concentrations of DMA and MMA stock standards as required in the method. In addition, the AOAC INTERNATIONAL Interlaboratory Study Workbook for Blind (Unpaired) Replicates was used to estimate repeatability and reproducibility of the method (18).

Analytical Limits

Before analyzing validation samples, laboratories were instructed to estimate their analytical solution detection limit (ASDL), analytical solution quantitation limit (ASQL), and corresponding method LOD and LOQ for As(III), As(V), DMA, and MMA in RTD juice and juice concentrate samples. These results are tabulated in Table 2. Laboratories determined the concentration SDs (σ) for each arsenic species by analyzing 10 replicates of a low-level mixed standard containing between 0.04 and 0.4 ng/g each of As(III), As(V), DMA, and MMA. Per the FDA EAM 3.2, the ASDL and ASQL values for each arsenic species were calculated using approximately 3.8σ and 30σ, respectively (19). The use of 30σ to calculate ASQL, rather than the 10σ commonly used, accounts for uncertainty in the signal measurement used to determine ASDL, as well as additional sources of uncertainty. ASDL and ASQL results obtained for DMA and MMA by one laboratory were significantly higher compared to the other seven participants and were rejected as outliers using both the Dixon and Grubbs tests. The higher limits obtained by this laboratory were concluded to be related to a known issue with the verification of the DMA and MMA standard concentrations. As expected, the average ASDLs obtained for each of the species based on arsenic concentrations were very similar, with averages ranging from 0.043 μg/kg for As(III) to 0.047 μg/kg for As(V).

Table 2.

Summary of ASDL, ASQL, and respective method LOD and LOQ for iAs, DMA, and MMA (n = 7 laboratories)

Analytical limit	Sample	Parameter	As(III), μg/kg	As(V), μg/kg	iAs, μg/kga	DMA, n/g	MMA, ng/g
ASDL		n	8	8		7	7
		$\overline{\mathrm{x}}$	0.043	0.047		0.046	0.045
		σ	0.011	0.005		0.016	0.014
		RSD	26%	11%		35%	31%
ASQL		$\overline{\mathrm{x}}$	0.32	0.36		0.36	0.34
LOD	RTD juice	$\overline{\mathrm{x}}$	0.22	0.24	0.24	0.23	0.23
LOQ	RTD juice	$\overline{\mathrm{x}}$	1.6	1.8	1.8	1.8	1.7
LOD	Juice concentrate	$\overline{\mathrm{x}}$	1.3	1.4	1.4	1.4	1.4
LOQ	Juice concentrate	$\overline{\mathrm{x}}$	9.7	11	11	11	10

^aThe method LOD and LOQ for iAs assumes that at a minimum, either As(III) or As(V) must be detected at a level ≥LOD.

The respective method LODs and LOQs for As(III), DMA, MMA, and As(V) were calculated using a dilution factor of 5× for RTD juice samples and 30× for juice concentrate samples and are shown in Table 2. Because this method does not provide effective control over potential oxidation/reduction of iAs species during sample preparation, iAs results are provided as a sum of the As(III) and As(V) detected. This makes interpretation of the iAs method LOD more difficult. The method LOD for iAs was conservatively estimated as the highest method LOD among those obtained for As(III) and As(V) and assumes that, at a minimum, either As(III) or As(V) must be detected at a level equal to or greater than the iAs LOD. The average method LODs and LOQs for iAs in RTD juices and juice concentrates obtained are shown in Table 2. Based on these results, the typical analytical limits listed in AOAC First Action Official Method 2016.04 are estimated as follows: method LOD = 0.25 μg/kg and LOQ = 2.0 μg/kg for iAs, DMA, and MMA in RTD juices; method LOD = 1.5 μg/kg and LOQ = 12 μg/kg for iAs, DMA, and MMA in juice concentrates.

Calibration and Method Blanks

Laboratories completed this study by running three to five analytical batches. Calibration curves were generated for each of these batches over the working range. Typically, four to five calibration levels over the range of 0.5–20 ng/g were used; R² values reported were >0.995 in all but one batch for two analytes from one laboratory. Calibration check standards, typically a 2 or 4 ng/g mix standard, were analyzed by all eight laboratories periodically throughout each analytical batch. The measured solution concentrations were monitored and compared to their prepared concentrations (reported as percent recoveries). For a total of 124 check standard determinations, the averages ±1 SD (and ranges) were as follows: iAs, 102 ± 4% (88–111%); DMA, 101 ± 5% (88–113%); and MMA 102 ± 5% (85–113%). Retention times (t_R) for check standards varied within an analytical batch for As(III), DMA, MMA, and As(V) by an average (and maximum) of 0.03 min (0.07 min), 0.04 min (0.13 min), 0.08 min (0.27 min), and 0.19 min (0.35 min), respectively (data not shown). Over time and multiple batches analyzed, MMA and As(V), in particular, will show larger variability with t_R, generally shortening as columns foul.

Six to 10 method blanks were analyzed by each laboratory over the course of this study (n = 66); and none of the four analytes was detected above the calculated LODs in 65 of 66 method blanks. A small amount of As(V) attributed to an autosampler carryover issue was detected in one method blank. These results were expected based on the relatively simple sample preparation process and the ability to easily control contamination in those steps. However, during preliminary work before the validation exercise, some laboratories reported As(V) contamination in method blanks. This contamination was found to originate from the (NH₄)₂HPO₄ used to prepare the mobile phase used. The contamination was apparent when the baseline of the m/z 75 trace would noticeably dip at the void volume of the separation, near AsB elution, and As(V) would be detected. This is surmised to occur when As(V) collected on the column head or within the injector and was released periodically, thus resulting in a small As(V) peak. A protocol was developed to determine whether the mobile phase salt being used contained an unacceptable level of arsenic. This is described in First Action Official Method 2016.04 [section I(d)(1)–(5)]. For the sake of efficiency, the central laboratory tested multiple (NH₄)₂HPO₄ sources and supplied approximately 40 g from an acceptable source to each laboratory along with the validation samples.

Laboratories were also instructed to run method blank fortifications at two spike levels, including one level near each laboratory’s LOQ for RTD juices (approximately 0.4 ng/g in the analyzed solution, equivalent to 2 μg/kg in the sample) and another level equivalent to 25 μg/kg. Seven laboratories prepared low-level method blank fortifications, with five laboratories using spike levels of 2 μg/kg and the other two laboratories using levels of 5 and 10 μg/kg. Each laboratory prepared between three and six replicates for a total of 31 replicates. The average recoveries ±1 SD of the low-level-fortified method blanks were 100 ± 8, 97 ± 9, and 101 ± 8% for iAs, DMA, and MMA, respectively. A total of three individual fortification recoveries [one each for iAs (79%), DMA (69%), and MMA (64%)] were outside the 80–120% acceptable range set forth in the QC criteria for the method. Both of the samples that yielded recoveries <70% were prepared by the same laboratory. Six laboratories prepared 25 μg/kg method blank fortifications, and one laboratory prepared their high-level method blank fortification at a concentration of 10 μg/kg. Each laboratory prepared between three and eight replicates for a total of 30 replicates. The average recoveries ±1 SD of the high-level-fortified method blanks were 102 ± 5, 100 ± 6, and 101 ± 6% for iAs, DMA, and MMA, respectively, and no recoveries were outside the 80–120% acceptable range.

Results for RTD Juice and Juice Concentrate Validation Samples

Reported concentrations of iAs, DMA, and MMA from each laboratory are summarized in Figure 1 and Table 1. Values between the method LOD and LOQ are reported as “trace.” As discussed earlier, the results from one laboratory for DMA and MMA were rejected for a failure to verify DMA and MMA standard concentrations as directed in the method. In addition, one replicate of grape juice sample (GJ2) from one laboratory was rejected as an outlier by the Cochran test. The values for iAs and DMA in the rejected replicate were approximately 50% of the overall average values.

Figure 1.

Speciation results for RTD and concentration juice samples from each laboratory. (a) RTD apple juice; (b) apple juice concentrate; (c) RTD grape juice; and (d) RTD pear juice.

Repeatability SD (s_r), reproducibility SD (s_R), RSD_r, RSD_R, and Horwitz ratio (HorRat) values were calculated for all average results found to be above the method LOD, including those that were considered trace levels, and are presented in Table 1. Eleven of the 13 samples had average iAs concentrations that were above the method LOQs; for RTD juices, these results covered the concentration range of approximately 5–25 μg/kg. For these 11 samples, the RSD_r ranged from 2.1 to 5.2%, the RSD_R ranged from 3.9 to 12%, and the HorRat values ranged from 0.13 to 0.38. The two samples that were found to have average iAs concentrations <LOQ were both juice concentrates. One of the two samples, AJC2, had an average iAs concentration of 9.94 μg/kg, just below the LOQ of 11 μg/kg. For this sample, the RSD_R and HorRat values were still reasonable at 27% and 0.85, respectively. These results indicate excellent within-laboratory and among-laboratory precision for iAs determinations using this method.

As discussed previously, As(III) and As(V) are determined individually and summed to report iAs concentration because of potential oxidation/reduction of iAs species during sample preparation. Evidence for the need to do this is provided when looking at results obtained for As(III), As(V), and iAs in two grape juice samples, GJ2 and GJ3 (individual data for As(III) and As(V) not shown). In both of these samples, the determined As(III) and As(V) concentrations were roughly equal, ranging from 5.3 to 8.5 μg/kg. In sample GJ2, the RSD_R values for As(III) and As(V) were 6.3 and 6.1%, respectively, whereas the iAs RSD_R was only 3.9%. Similarly, for sample GJ3, the RSD_R values for As(III) and As(V) were 7.6 and 8.5%, respectively, whereas the iAs RSD_R was 4.9%.

One of the 13 samples was fortified by the originating laboratory at a level of 30 μg/kg DMA. All other samples were not found to contain DMA at a concentration above the method LOQ. The RSD_r, RSD_R, and HorRat values for this sample were 3.3%, 6.7%, and 0.24, respectively. Four RTD juice samples contained trace levels of DMA that were >2 times the method LOD. For these samples, the RSD_R ranged from 28 to 33% and the HorRat values ranged from 0.60 to 0.73.

Four of the 13 samples were found to have MMA concentrations above the method LOQ; however, three of those were fortified by the originating laboratory. For these samples, the MMA RSD_r values ranged from 2.2 to 4.4% and the RSD_R values ranged from 6.0 to 9.3%. The HorRat values were all less than 0.5. One sample, AJ3, contained a trace level of MMA that was >2 times the method LOD. For this sample, the RSD_R and HorRat values were 58% and 1.24.

Sample Fortification Recovery

Each laboratory participating in this MLV study prepared fortified samples representing each matrix (AJ1, GJ1, PJ1, and AJC1) in duplicate at three spike concentration levels and then determined spike recoveries. The nominal spike concentrations in the RTD juices were 5, 25, and 50 μg/kg for As(III), As(V), DMA, and MMA. Thus, in calculating the recovery for iAs, the fortification concentrations and determined concentrations for both As(III) and As(V) were taken into account. The results of this fortification study are shown in Table 3. Overall, fortification recoveries in RTD juice samples of apple, grape, and pear were very good, with averages ±1 SD of 100 ± 8, 100 ± 7, and 100 ± 8% for iAs (n = 144), DMA (n = 126), and MMA (n = 126), respectively. Only one recovery values for iAs and one recovery values for MMA were outside the 80–120% acceptable range set forth in the QC criteria for the method, and each of these values was >120% but ≤130% recovery. Again, due to a known standard issue, the data for DMA and MMA for one laboratory were rejected from these results.

Table 3.

Average, SD, minimum, and maximum spike recoveries in RTD apple, grape, and pear juices and apple juice concentrate

Analyte	Matrix	No. of Labs	No. of replicates	Nominal spike level, μg/kg	Average spike recovery, %	SD spike recovery, %	Minimum spike recovery, %	Maximum spike recovery, %
iAsa	AJ1	8	16	5 As(III) + 5 As(V)	104	11	88	125
	AJ1	8	16	25 As(III) + 25 As(V)	104	7	91	117
	AJ1	8	16	50 As(III) + 50 As(V)	102	9	83	119
	GJ1	8	16	5 As(III) + 5 As(V)	101	8	86	115
	GJ1	8	16	25 As(III) + 25 As(V)	97	6	81	106
	GJ1	8	16	50 As(III) + 50 As(V)	98	4	89	106
	PJ1	8	16	5 As(III) + 5 As(V)	97	8	83	109
	PJ1	8	16	25 As(III) + 25 As(V)	98	6	88	106
	PJ1	8	16	50 As(III) + 50 As(V)	97	6	88	108
	AJC1	8	16	35 As(III) + 35 As(V)	105	9	89	119
	AJC1	8	16	100 As(III) + 100 As(V)	103	10	85	119
	AJC1	8	16	200 As(III) + 200 As(V)	104	8	91	120
DMA	AJ1	7	14	5 DMA	102	10	86	119
	AJ1	7	14	25 DMA	102	6	92	115
	AJ1	7	14	50 DMA	101	6	89	115
	GJ1	7	14	5 DMA	99	6	90	110
	GJ1	7	14	25 DMA	97	6	82	107
	GJ1	7	14	50 DMA	98	5	89	104
	PJ1	7	14	5 DMA	101	8	91	120
	PJ1	7	14	25 DMA	103	6	96	117
	PJ1	7	14	50 DMA	102	7	92	117
	AJC1	7	14	35 DMA	100	5	94	110
	AJC1	7	14	100 DMA	97	7	82	105
	AJC1	7	14	200 DMA	98	7	82	106
MMA	AJ1	7	14	5 MMA	101	9	82	114
	AJ1	7	14	25 MMA	101	5	94	110
	AJ1	7	14	50 MMA	100	6	89	109
	GJ1	7	14	5 MMA	105	8	89	118
	GJ1	7	14	25 MMA	96	8	82	105
	GJ1	7	14	50 MMA	98	7	84	110
	PJ1	7	14	5 MMA	98	15	80	130
	PJ1	7	14	25 MMA	101	8	89	116
	PJ1	7	14	50 MMA	100	5	90	108
	AJC1	7	14	35 MMA	100	7	89	113
	AJC1	7	14	100 MMA	97	9	80	109
	AJC1	7	14	200 MMA	97	6	87	105

^aMeasured as As(III) + As(V).

The nominal spike concentrations in the apple juice concentrate were higher than those for RTD juices because of the additional sample dilution that is required before analysis. The nominal spike concentrations were 35, 100, and 200 μg/kg for As(III), As(V), DMA, and MMA. Overall, fortification recoveries in apple juice concentrate were very good, with averages ±1 SD of 104 ± 9, 98 ± 7, and 98 ± 7% for iAs (n = 48), DMA (n = 42), and MMA (n = 42), respectively. None of the recoveries were outside the 80–120% acceptable range. The results in Table 3 for the RTD juices and the apple juice concentrate do not appear to show any significant differences in recoveries based on juice type and/or fortification level. These fortification recovery results for apple, grape, and pear RTD juices and apple juice concentrate demonstrate acceptable accuracy and precision of the method over the range of concentrations of interest and do not indicate any significant matrix interferences, including signal enhancement for iAs, DMA, and MMA, in these sample types.

Sample fortifications provided a good proxy for t_R assessment of standard versus spiked sample. In general, As(V) and MMA showed the largest time shift [As(III) and DMA changes were insignificant, data not shown]. The maximum differences observed within a batch were 0.9 min for MMA and 0.6 min for As(V). These shifts were most likely reflective of the pH difference between the standard and sample solutions. The MMA deviation was more pronounced because the pKa₂ of MMA is 8.7 (20), near the pH of the mobile phase (8.2–8.3).

Mass Balance

Total arsenic values of all 13 juice samples were determined by two participating laboratories using the FDA method EAM 4.7 (21). These values were used to calculate the respective mass balances between the sum of the arsenic species determined and the total arsenic determined in the samples. Good mass balance indicates that the majority of the total arsenic in the samples is accounted for during the speciation analysis. Total arsenic values were not provided to the laboratories before starting the exercise. As shown in Table 4, the mass balances obtained for all 13 samples were within the method’s acceptable range of 65–115%. Two RTD juice samples with total arsenic ≤10 μg/kg and two apple juice concentrate samples with total arsenic <15 μg/kg had mass balances <70%, whereas all other samples had mass balances ranging from 84 to 110%. The speciation results for the two juice concentrate samples with lower mass balances were <LOQ. The lower mass balances for the RTD grape juice sample GJ1 and pear juice sample PJ3 are not as easily explained because the iAs determined in both of these samples was approximately 3 times the method LOQ. In GJ1, a trace amount of DMA was determined. The higher uncertainty associated with trace determinations and the relatively low concentration of iAs in the sample most likely contributed to the lower mass balance obtained.

Table 4.

Summary of mass balance data for RTD apple, grape, and pear juice and apple juice concentrate samples

Sample	Analyte added, μg/kg	Sum of As species concn, μg/kga	Total As value, μg/kgb	Mass balance, %
AJ1	NAc	11.5 ± 0.8	13.4 ± 1.9	86
AJ2	25 As(V), 25 MMA	47.3 ± 4.4	51.9 ± 5.5	91
AJ3	NA	6.4 ± 0.6	7.6 ± 0.88	84
GJ1	NA	6.9 ± 0.5	10.2 ± 1.7	68
GJ2	30 DMA	43.2 ± 1.6	46.9 ± 2.6	92
GJ3	NA	13.5 ± 0.8	12.3 ± 2.5	110
GJ4	10 As(V), 10 MMA	33.1 ± 1.7	37.3 ± 3.2	89
PJ1	30 MMA	33.9 ± 2.0	34.1 ± 2.9	99
PJ2	10 As(III)	15.1 ± 1.2	16.6 ± 0.6	91
PJ3	NA	5.8 ± 0.8	8.8 ± 0.28	66
AJC1	NA	87.0 ± 5.2	97.5 ± 3.5	89
AJC2	NA	9.9 ± 3.3	14.7 ± 2.3	67
AJC3	NA	4.9 ± 3.2	7.2 ± 1.68	68

^aSum of As species = iAs + DMA + MMA; mean ± 1 σ with seven laboratories reporting (n = 21).

^bTotal As value from ICP-MS analysis by two laboratories; mean ± 1 σ.

^cNA =Not applicable.

Analysis of Certified Reference Materials

It is unfortunate that Certified Reference Materials for arsenic species in fruit juice were not available at the time of this study. NIST SRM 1643e Trace Elements in Water, which is certified for total arsenic concentration at 58.98 ± 0.70 μg/kg, was analyzed by seven laboratories. In addition, two laboratories analyzed NIST SRM 1640a Trace Elements in Natural Water, which is certified for total arsenic concentration at 8.010 ± 0.067 μg/kg. Each laboratory analyzed a minimum of one preparation of SRM with each batch of samples analyzed. One concern regarding the use of either of these reference materials was the presence of nitric acid (5%, v/v) in the standards and its potential effect on the chromatography. Dilution of NIST SRM 1643e using water by a factor of 7.5 results in a nitric acid concentration of <1%, which did not significantly affect the chromatography. Smaller dilutions resulting in higher concentrations of nitric acid noticeably decreased the t_R of the As(V) peak (data not shown).

Results for the analysis of reference materials are shown in Table 5. The overall average concentration of iAs determined in SRM 1643e was 56.7 ± 5.1 μg/kg (n = 33), which accounts for 96% of the certified value for total arsenic. For both MMA and DMA, 32 of 33 replicates were found to be <LOD. The results for one replicate of DMA and one replicate of MMA from different laboratories were rejected as outliers and likely resulted from contamination during sample preparation or carryover from a previous injection. Two laboratories analyzed SRM 1640a (one exclusively), and the average concentration of iAs determined was 7.89 ± 0.94 μg/kg (n = 10), accounting for 98% of the certified value for total arsenic. The DMA and MMA results for all 10 replicates of SRM 1640a were <LOD. In addition, all 43 determinations of iAs, including both SRM 1643e and 1640a, were within the total certified arsenic levels ±20%, QC control limit. These recoveries were within the acceptable range of 80–120% set forth in the QC criteria for the method and demonstrate good accuracy of the method for iAs determination.

Table 5.

Analysis of Certified Reference Materials

Reference Material	No. of Labs	No. of replicates	iAs, μg/kga	DMA, μg/kg	MMA, μg/kg	Sum of As species, μg/kga	Total As certified value, μg/kg	Total As control limit, μg/kg ± 20%b
NIST SRM 1643e Trace Elements in Water	7	33	56.7 ± 5.1	<LOD	<LOD	56.7 ± 5.1	58.98 ± 0.70	59.0 ± 12.0
NIST SRM 1640a Trace Elements in Natural Water	2	10	7.89 ± 0.94	<LOD	<LOD	7.89 ± 0.94	8.010 ± 0.067	8.01 ± 1.60

^amean ± 1 σ.

^bUncertainty expressed as ± 20% of the certified values.

Conclusions

This method has provided acceptable results at and below the level of interest (10 μg/kg) for iAs, DMA, and MMA in the matrixes studied. Accuracy and trueness of the method were demonstrated through fortification recovery studies at three concentration levels in four matrix types. Overall, the results were excellent, with average recoveries ranging from 96 to 105% across all analytes. The RSDs obtained in fortification recovery studies were generally <15%. In addition, results for iAs in NIST SRMs 1643e and 1640a were in excellent agreement, with recoveries of 96–98% when compared with the certified total arsenic concentrations. Precision within and among laboratories was determined through the analysis of 13 samples, including RTD apple, grape, and pear juices and apple juice concentrates. For analytes above the method LOQ, the reproducibility was <15%.