Speciated arsenic in air: Measurement methodology and risk assessment considerations

Abstract

Accurate measurement of arsenic (As) in air is critical to providing a more robust understanding of arsenic exposures and associated human health risks. Although there is extensive information available on total arsenic in air, less is known on the relative contribution of each arsenic species. To address this data gap, the authors conducted an in-depth review of available information on speciated arsenic in air. The evaluation included the type of species measured and the relative abundance, as well as an analysis of the limitations of current analytical methods. Despite inherent differences in the procedures, most techniques effectively separated arsenic species in the air samples. Common analytical techniques such as inductively coupled plasma mass spectrometry (ICP-MS) and/or hydride generation (HG)- or quartz furnace (GF)-atomic absorption spectrometry (AAS) were used for arsenic measurement in the extracts, and provided some of the most sensitive detection limits. The current analysis demonstrated that, despite limited comparability among studies due to differences in seasonal factors, study duration, sample collection methods, and analytical methods, research conducted to date is adequate to show that arsenic in air is mainly in the inorganic form. Reported average concentrations of As(III) and As(V) ranged up to 7.4 and 10.4 ng/m⊃, respectively, with As(V) being more prevalent than As(III) in most studies. Concentrations of the organic methylated arsenic compounds are negligible (in the pg/m3 range). However, because of the variability in study methods and measurement methodology, the authors were unable to determine the variation in arsenic composition as a function of source or particulate matter (PM) fraction. In this work, the authors include the implications of arsenic speciation in air on potential exposure and risks. The authors conclude that it is important to synchronize sample collection, preparation, and analytical techniques in order to generate data more useful for arsenic inhalation risk assessment, and a more robust documentation of quality assurance/quality control (QA/QC) protocols is necessary to ensure accuracy, precision, representativeness, and comparability.

Implications:

The detailed review of the speciation techniques for arsenic in air samples presented in this study showed that (1) a variety of efficient methods (sampling, extraction, and analytical) developed across research studies were available for the determination of arsenic species in air; (2) arsenic species in the air were predominantly inorganic forms; and (3) the organic forms in air were largely below detection limits. Despite these general findings, more consistent methods must be developed to understand potential risks from arsenic in ambient air. Research needs to advance air arsenic risk assessment are identified.

Introduction

The study of arsenic in environmental media (i.e., groundwater, soil, and air) and the associated human health risks has been an active area of research for many decades. Due to the relatively high concentrations of naturally occurring arsenic in groundwater in many parts of the world, the relationship between arsenic in drinking water and adverse health effects has been studied extensively. Environmental exposure to arsenic in soil has also been the subject of substantial research, mainly in the context of population exposures to nearby mining and smelting activities. Conversely, the potential for health risks associated with exposure to arsenic occurring in ambient air remain less understood, and generally restricted to occupational, rather than environmental, health studies.

Traditional arsenic risk assessment has relied heavily on evaluating exposure to total environmental arsenic, without regard to chemical speciation. For example, regulatory criteria (e.g., maximum contaminant levels [MCLs], soil screening levels [SSLs]) are based on total arsenic measurements. Arsenic, however, occurs in many different forms, with each exhibiting unique toxicological properties. In recent years, research efforts have focused on understanding the health risks associated with specific forms of arsenic. Insight into these differences, in combination with the potential implications for arsenic regulation, has in turn spurred the development of analytical methods capable of differentiating arsenic species in various environmental media.

In keeping pace with health-related research, measurement method development has been largely directed toward understanding arsenic speciation in water, soil, and food matrices, as well as biological substrates (e.g., urine, hair, nails). Less research has been directed towards differentiating between arsenic species in air. This paper provides a comprehensive summary of peer-reviewed information on speciated arsenic in air, including an evaluation of the species measured and their relative abundance by arsenic type. A key aim of this paper is to understand the current state-of-the-art analytical methods for measuring speciated arsenic in air. Additionally, this analysis provides an evaluation of sample preparation and measurement techniques across studies, as well as identifies potential deficiencies in current analytical methods. Available information is evaluated in the context of identifying research needs that will better inform the evaluation of human health risks related to arsenic in air.

Methods

This literature review involved searching the peer-reviewed literature using the PubMed (http://www.ncbi.nlm.nih.gov/pubmed/), Dialog (http://www.dialogclassic.com/), and Scopus (http://www.scopus.com) search engines to identify studies that measured speciated arsenic in air. The Dialog search included the ToxFile, EMBASE, MEDLINE, AGRICOLA, Ei Compendex, Environmental Sciences, Enviroline, and SciSearch databases. Species of arsenic of potential interest include trivalent inorganic arsenic (As(III)), pentavalent inorganic arsenic (As(V)), monomethylarsonic acid (MMA), and dimethylarsinic acid (DMA). Measures of total inorganic arsenic and total organic arsenic were also considered of interest.

Relevant air studies were identified using the search terms listed below, and the search was limited to documents published in English.

1.	“Arsenicals” OR “Cacodylic Acid” OR arsenates OR arsenites OR roxarsone OR arsine OR arsonic OR msma OR “Arsanilic Acid” OR arsenic
2.	Speciated OR speciation OR species OR methylated OR methylation OR trivalent
3.	Air OR particulate OR TSP OR PM10 OR PM2.5 OR aerosol OR atmosphere OR emission OR particle(s)

All titles returned based on searches performed with terms 1, 2, and 3 were considered, and the abstracts of each potentially relevant title were then reviewed to identify the studies germane to this specific analysis (i.e., containing at least one speciated arsenic result). In papers identified as containing speciated arsenic measurements, results for speciated and total As, sample location, particulate size fraction, sample collection methods, sample extraction and analysis methods, and quality assurance/quality control (QA/QC) results were tabulated.

Results and Discussion

Speciated arsenic concentrations in air

Specific study results

The literature search returned 12 relevant publications reporting arsenic species measurements in ambient air. These studies measured speciated arsenic in particulate matter (PM), including PM₁₀, PM_2.5, and/or total suspended particles (TSP) from sampling locations around the world. The studies reported concentrations in predominantly urban and industrial locations, with more limited sampling of rural locations. The key results are summarized in Table 1 (results are separated out by particle size).

Table 1. Mean (range) of speciated arsenic concentrations in ambient air (ng/m³), grouped by particle size

Study(Reference)	ParticleSize	EnvironmentalSetting	Numberof Samples	As(III)	As(V)	Total As	DMA	MMA	TMA^g	As(III)/As(V)	(As(III) + As(V))/Total
Speciated arsenic in PM2.5
Šlejkovec et al., 2000	<2 µm	U; D	2	ND (<0.01–<0.02)	1.18 (0.84–1.52)	2.4 (2.1–2.6)	ND (<0.05–<0.10)	ND (<0.01–<0.01)			0.49
Šlejkovec et al., 2000	<2 µm	U; AT	2	ND (<0.05–<0.09)	0.21(0.12–0.3)	0.47 (<0.3–0.64)	ND (<0.26–<0.47)	ND (<0.02–<0.03)			0.45
Rabano et al., 1989	<2.5 µm	U	16	7.4 (<1.2–44)	5.2 (<0.9–18.7)					1.42
Sánchez de la Campa et al., 2008	<2.5 µm	U; I	22	1.4 (0.1–2.7)	6.6 (0.01–56.2)	7.9 (1–56.6)				0.21	1.01
Sánchez de la Campa et al., 2008	<2.5 µm	U; I	24	0.9 (0.01–1.6)	5 (0.01–25.3)	6.4 (0.8–30.2)				0.18	0.92
Farinha et al., 2004	<2.5 µm	U; I^a	47	0.09 (NR)	1.14 (NR)	2.71 (NR)				0.079	0.45
Farinha et al., 2004	<2.5 µm	U; I^b	43	0.037 (NR)	0.42 (NR)	1.9 (NR)				0.088	0.24
Tsopelas et al., 2008	<2.5 µm	I^c	16	<0.2 (NR)	1.7 (NR)	1.9 (NR)	<0.5 (NR)	<0.5 (NR)			0.89
Range of mean arsenic concentration in PM2.5				ND-7.4	ND-6.6	0.047–7.9				0.000–1.42	0.24–1.01
Speciated arsenic in PM10
Sánchez-Rodas et al., 2007	<10 µm	U; I	25	1.2 (0.6–2.2)	6.5 (0.01–25.7)	7.7 (1.6–29.4)				0.18	1.00
Sánchez-Rodas et al., 2007	<10 µm	U; I	25	2.1 (0.4–3.4)	7.8 (0.01–77.2)	9.9 (1.3–79.8)				0.27	1.00
Range of mean arsenic concentration in PM10				1.2–2.1	6.5–7.8	7.7–9.9				0.18–0.27	1.00–1.00
Speciated arsenic in coarse PM
Šlejkovec et al., 2000	2–10 µm	U; D	2	ND (<0.01–<0.02)	0.115 (0.08–0.15)	0.96 (0.61–1.3)	ND (<0.05–<0.1)	ND (<0.01–<0.01)			0.12
Šlejkovec et al., 2000	2–10 µm	U; AT	2	ND (<0.09–<0.11)	0.545 (0.41–0.68)	4.5 (3.7–5.3)	ND (<0.47–<0.58)	ND (<0.03–<0.04)			0.12
Farinha et al., 2004	2.5–10 µm	U; I^a	47	0.032 (NR)	0.55 (NR)	0.69 (NR)				0.058	0.84
Farinha et al., 2004	2.5–10 µm	U; I^b	43	0.025 (NR)	^f(NR)	0.15 (NR)					0.17
Tsopelas et al., 2008	2.5–10 µm	I^c	16	<0.2 (NR)	1.0 (NR)	1.1 (NR)	<0.5 (NR)	<0.5 (NR)			0.91
Rabano et al., 1989	>2.5 µm	U	16	1.8 (<0.9–4.8)	2.2 (<0.8–6.6)					0.82
Range of mean arsenic concentration in coarse PM				ND-1.8	0.1–2.2	0.15–4.5				0.00–0.82	0.12–1.00
Speciated arsenic in TSP
Rabano et al., 1989	TSP	U	16			18.9 (2.8–69.4)
Oliveira et al., 2005	TSP	U; I	12	1.2 (0.3–1.8)	10.4 (2.1–30.6)	12.3 (3.0–33.8)	ND (NR)	ND (NR)		0.12	0.94
Solomon et al., 1993	TSP	R; I^d	17	0.54 (0.06–3.08)	2.29 (0.51–6.54)					0.24
Solomon et al., 1993	TSP	Rl; I	19	0.22 (<0.18–<0.3; max detect 0.19)	0.46 (0.08–1.3)					0.48
Mukai and Ambe, 1987b	TSP	R	1				0.017	ND	0.057
Mukai and Ambe, 1987b	TSP	R	1				0.008	ND	0.006
Mukai and Ambe, 1987b	TSP	R	1				0.003	ND	0.02
Mukai and Ambe, 1987b	TSP	R	5				0.025 (0.006–0.053)	0.097 (ND-0.485)	0.177 (0.025–0.604)
Mukai and Ambe, 1987a	TSP	R; AF^e	1				NR (NR-0.43)	NR (NR-1.4)
Tsopelas et al., 2008	TSP	I^c	16	<0.2 (NR)	3.2	3.4	<0.5 (NR)	0.5 (NR)			0.94
Range of mean arsenic concentration in TSP				0.22–1.2	0.46–10.4	12.3–18.9				0.12–0.48	0.94–0.94
Notes: AF = agricultural fields; AT = automobile tunnel; D = downtown; I = industrial; ND = below detection limit; results below ND were replaced with DL in order to calculate averages; NR = not reported; R = rural; U = urban. ^aLocated near a 100-MW coal/oil power plant. ^bLocated near a 1256-MW coal power plant. ^cLocated near refineries and a major highway. ^dLocated in a geothermal development area. ^eLocated 20–30 km away from an area where alkyl-arsenic pesticides were applied in June and early July. ^f Farinha et al. (2004), reported 0.005 ng/m3 of As(V) in the coarse fraction from Sines. This result was based on a single measurement, and it is inconsistent with all other As(V) results in this study (i.e., As(V) concentrations are expected to be higher than As(III) concentrations). Therefore, we have not reported the As(III):As(V) ratio for this result. ^gTrimethyl arsenic.

In one of the first U.S. studies to evaluate speciated arsenic in atmospheric samples, Johnson and Braman (1975) measured speciated arsenic (methylarsine and dimethylarsine) in PM greater than and less than 3 µm in indoor and outdoor air samples collected in urban and suburban residential areas in Florida. In outdoor samples, the average total arsenic concentration was 1.7 ng/m3, with 20% in the organic (methyl-, dimethyl-, trimethylarsines) form, although the authors reported significant variability. Approximately half of the organic arsenic was found in each of the size fractions. In indoor samples, the average total arsenic concentration was 8 ng/m3, with a larger fraction of inorganic arsenic. The authors were unable to distinguish between the arsines and organic acids, thus the utility and comparability of this study to the others studies that distinguished between arsenic and other organic arsenic is quite limited.

Just over a decade later, prompted by concerns regarding the contamination of air from the use of organic arsenic pesticides in Japan, Mukai and Ambe developed an analytical method to measure methyl arsenic (mono-, di-, and tri-) species in airborne PM (Mukai and Ambe, 1987b) The method was applied to ambient particulates collected from four rural locations in Japan (Tsukuba City, Oorai, Katsuura, and Iioka) in the summer of 1985 and the spring of 1986 (Mukai and Ambe, 1987b) Total MMA and DMA concentrations in the PM ranged from nondetected (ND) to 485 and 3–53 pg/m3, respectively, with higher concentrations in July and August than in February and March. A second report described analysis of samples collected from an additional rural location 20–30 km from agricultural fields previously treated with arsenical pesticides (Mukai and Ambe, 1987a) The maximum total MMA and DMA concentrations were 1.4 and 0.43 pg/m3, respectively. Results are summarized in Table 1. Both studies also measured trimethyl arsenate (TMA); unfortunately, as these are the only studies reporting on this chemical species, no further discussion of TMA is included in this review.

In an effort to evaluate the potential toxicity of different arsenic species, Rabano et al. (1989), developed sampling, extraction, and analysis procedures for total and inorganic arsenic species in airborne PM. This study was performed in response to California's 1985 Air Toxics Law and represents some of the first analytical results for the individual species of As(III) and As(V) in air ever published. The samples were collected in Los Angeles County, CA, during January and February of 1987. They found an average concentration of 19.9 ng/m³ total arsenic in TSP. Inorganic arsenic concentrations were higher in PM_2.5 than in the coarse fraction (particles >2.5 µm) with average As(III) concentrations of 7.4 and 1.8 ng/m3, respectively, and average As(V) concentrations of 5.2 and 2.2 ng/m³, respectively.

In a related California study, Solomon et al. (1993), measured inorganic arsenic species in airborne PM collected from two sites in the Geysers Geothermal Development area (located approximately 70 miles north of San Francisco, CA) during 4 weeks in April and May of 1989. As geothermal steam is a natural source of arsenic, such areas could influence the concentrations of atmospheric arsenic in California. The study reported average As(III) concentrations of 0.54 ng/m³ (eastern sampling location) and 0.22 ng/m⊃ (western sampling location) from two sites within the study area. As(V) concentrations were higher, with 2.29 ng/m³ at the eastern location and 0.46 ng/m³ at the western location. The authors determined that the averages reported were consistent with the statewide average reported by the California Air Resources Board (ARB) of 1.5 ng/m³ (with averages of 1.2 and 2.0 ng/m³ in northern and southern California, respectively).

Talebi (1999), measured total arsenic and As(III) in PM collected in the city of Ifshan, Iran. As(III) constituted more than 50% of the total arsenic in PM. However, the lack of detailed results hinders additional interstudy comparison.

Šlejkovec et al. (2000), developed a sequential extraction method for measuring total and inorganic arsenic species in airborne PM collected in May 1998 from Castle District Tunnel and Széna Square, two urban sites in Budapest, Hungary. Total arsenic ranged from 0.47 to 4.5 ng/m³, with maximum concentrations varying with particle fraction between the two locations—highest in the coarse fraction (2–10 µm) in Castle Tunnel versus the fine fraction (<2 µm) in the Széna Square. By species, only As(V) was detected in the PM, with concentrations of 1.18 ng/m⊃ in the fine fraction and 0.12 in ng/m⊃ in the coarse fraction from Széna Square. In Castle Tunnel, As(V) was detected with concentrations of 0.21 ng/m³ in the fine fraction and 0.55 ng/m³ in the coarse fraction. All other species (i.e., As(III), MMA, and DMA) were not detected; as suggested by the authors, underestimation of As(III) may have occurred due to oxidation during sample preparation (Šlejkovec et al., 2000).

Farinha et al. (2004), measured total and inorganic arsenic species in fine (<2.5 µm) and coarse (2.5–10 µm) fractions of ambient PM, as well as lichens; the authors were interested in determining whether the lichens were biotransforming inorganic arsenic in atmospheric particulate to organic As. Samples were collected in 1994 and 1995 from two towns in Portugal, Tapada do Outerio and Sines, both located near coal-fired power plants. As(III) ranged from 0.025 to 0.09 ng/m³, and As(V) ranged from 0.42 to 1.14 ng/m³ (excludes the Sines As(V) measurement based only on a single value; all other results were an average of 40 values). The results indicated that total arsenic and the inorganic species were higher in the fine fraction, which the authors noted contrasted with the results of Šlejkovec et al. (2000), potentially due to different compositions of atmospheric particles in both countries (Hungary and Portugal) or the potential influence of Sahara-derived dust.

In 2004, the European Union (EU) published a directive that established a target value for total arsenic in ambient air of 6 ng/m³ in PM₁₀ averaged over a calendar year (European Parliament, European Union Council, 2005). Prompted by such evolving health-based guidelines regarding arsenic in air and the recognition that As(III) and As(V) differ in toxicity, three European studies evaluated speciated arsenic concentrations in Spain. First, Oliveira et al. (2005), measured total and speciated arsenic in 12 samples (1 per month) of TSP collected in Huelva, Spain, during 2000. They investigated several extraction methods to identify the optimal procedure. Annual average total arsenic was 12.3 ng/m³, whereas annual mean As(III) and As(V) were 1.2 and 10.4 ng/m³, respectively, when using NH₂OH·HCl + microwave digestion, and 0.9 and 10.1 ng/m³, respectively, when using H₃PO₄ + microwave digestion. Organic arsenic compounds were not detected. The study results indicated that a nearby copper smelter likely influenced the arsenic concentrations in the city of Huelva. Two later studies, Sánchez-Rodas et al. (2007), and Sánchez de la Campa et al., (2008), applied the technique developed by Oliveira et al. (2005), to measure total and speciated arsenic in PM₁₀ and PM_2.5 samples, again collected in Huelva. The PM samples were collected every 2 weeks during 2001 and 2002. Mean concentrations in the 2001 PM₁₀ samples (N = 25) were slightly lower than the concentrations measured in 2002 (1.2, 6.5, and 7.7 ng/m³ for As(III), As(V), and total As, respectively, versus 2.1, 7.8, and 9.9 ng/m³ for As(III), As(V), and total As, respectively). Mean concentrations in the PM _2.5 fraction were similar, but again lower. Concentrations in the 2001 PM_2.5 samples (N = 25) were 0.9, 5.0, and 6.4 ng/m³ for As(III), As(V), and total As, respectively, as compared to 2002 PM_2.5 samples (N = 25) at 1.4, 6.6, and 7.9 ng/m³ for As(III), As(V), and total As, respectively. The authors noted that the rural background concentrations in Huelva were high in comparison to other rural and urban areas in Spain, emphasizing the need to continue arsenic speciation research.

Also in consideration of the EU's Directive, Tsopelas et al. (2008), measured total and speciated arsenic in coarse PM and PM_2.5 collected in Aspropygos, Greece, between December 2004 and June 2006. They developed and evaluated effective procedures for extraction of all four arsenic species (As(III), As(V), MMA, and DMA). As an industrial area located near refineries and a major highway, the Aspropygos area was selected for the potential concentration of urban arsenic sources. The average total arsenic concentration was 1.1 ng/m³ in PM_10–2.5 and 1.9 ng/m³ in PM_2.5. Only As(V) was detected in these samples; As(III), MMA, and DMA were not detected in either size fraction. The average As(V) concentrations were 1.0 ng/m³ in PM_10–2.5 and 1.7 ng/m³ in PM_2.5. Higher concentrations were reported for all species in the summer versus the winter, which the authors attributed to differences in meteorological influences (wind and rain).

Overall study summary

A great deal of variability was demonstrated in the ambient air studies reviewed, including seasonal differences in sampling dates and duration, environmental setting, sample collection methods, and analytical methods. Just 2 of the 12 studies reviewed were performed in the United States (Rabano et al., 1989; Solomon et al., 1993). The remainder were conducted in Japan (Mukai and Ambe, 1987a Mukai and Ambe, 1987b) Hungary, (Šlejkovec et al., 2000), Portugal (Farinha et al., 2004), Spain (Oliveira et al., 2005; Sanchez de la campa et al., 2008; Sanchez-Rodas et al., 2007), and Greece (Tsopelas et al., 2008). The sampling dates span three decades (1975–2008), during which sampling and analysis techniques were evolving.

Most of the studies reviewed measured air from urban and/or industrial areas. Table 1 summarizes the mean and range of speciated and total arsenic concentrations found in PM across all of the studies. Studies that measured both As(III) and As(V) reported average concentrations up to 7.4 and 10.4 ng/m³, respectively. The calculated As(III):As(V) ratios ranged from 0.06 to 1.42. The sum of As(III) and As(V) represented 12–100% of total As. In the three studies in which all four species of arsenic were analyzed, the organic arsenic compounds were not detected when measured in conjunction with inorganic arsenic species (Oliveira et al., 2005; Šlejkovec et al., 2000; Tsopelas et al., 2008).

Although large differences appear across the studies, the arsenic species detected in PM were typically inorganic (i.e., more than 90%). In two studies, inorganic arsenic constituted less than half of the total As. Šlejkovec et al. (2000), detected only As(V) (i.e., MMA, DMA, and As(III) were smaller than detection limit, but the authors noted that there may have been oxidation from As(III) to As(V)), whereas Farinha et al. (2004), found 17–84% inorganic arsenic in coarse PM and 24–45% inorganic arsenic in PM_2.5. However, this later study did not analyze for organic arsenic species, and the efficiency of the sequential extraction approach suggests possible oxidation of As(III) to As(V); therefore, it is not possible to determine the quantitative nature of the relative disparities.

Based on the studies published to date, arsenic in ambient air PM presents as a mixture of inorganic trivalent and pentavalent species, with the latter appearing more prevalent. Specifically, based on the calculated As(III): As(V) ratio, Rabano et al. (1989), reported the highest proportion of As(III) compared to As(V), with a ratio of 1.42 in PM_2.5 and 0.82 in the coarse fraction. These results are in strong contrast to all other studies, which found As(III) to be much lower than As(V). For example, Solomon et al. (1993), reported ratios of 0.24 to 0.48 in rural California. The studies from the City of Huelva, Spain, found similar ratios from 0.12 in TSP to 0.27 in PM₁₀ (Oliveira et al., 2005; Sanchez-Rodas et al., 2007). In Portugal, Farinha et al. (2004), reported an even lower fraction of As(III), ranging from 0.06 to 0.08. As discussed previously, As(III) was not detected in the studies from Hungary and Greece (Šlejkovec et al., 2000; Tsopelas et al., 2008). This is consistent with the trends reported in the Agency for Toxic Substances and Disease Registry's (ATSDR's) Toxicological Profile (Agency for Toxic Substances and Disease Registry, 2007) and U.S. Environmental Protection Agency's (EPA's) Health Assessment Document for Inorganic Arsenic, (U.S. Environmental Protection Agency, 1984), which state that, in general, arsenic is found in particulates with higher levels of As(V) than As (III).

There was not enough information to evaluate urban versus rural variation in arsenic species in air. The only measurements that potentially address urban versus rural variation are from the two California studies, where inorganic arsenic concentrations in urban areas (7.4 ng/m³ As(III) and 5.2 ng/m³ as As(V) in PM_2.5) near Los Angeles (Rabano et al., 1989) were higher than concentrations in a rural geothermal development area (0.54 ng/m³ As(III) and 2.3 ng/m³ as As(V)) (Solomon et al., 1993). However, a geothermal development area is not representative of other more typical rural locations.

In the five studies evaluating the methylated arsenic species (MMA and DMA), generally neither was detected in urban/industrial or rural environments. Total MMA and DMA concentrations in PM ranged from nondetectable (ND) to 485 pg/m³ and 3 to 45 pg/m³, respectively, for four rural locations in Japan (Mukai and Ambe, 1987b) In a subsequent Japanese study, Mukai and Ambe (1987a) reported detectable organic arsenic species 20–30 km away from an agricultural area utilizing arsenical pesticides, but the concentrations appeared very low (i.e., the maximum total MMA and DMA concentrations at 1.4 and 0.43 pg/m³, respectively).

Analytical Techniques

Numerous methods have been published for arsenic analysis (U.S. Environmental Protection Agency, 1994, 2007). Most of these methods are designed to analyze arsenic in its total or “elemental” form in routine matrices such as soil and groundwater. With regard to arsenic determinations in ambient air, the EPA has published sampling and analysis procedures for metals in its Compendium of Methods for the Determination of Inorganic Compounds in Ambient Air. (EPA, 1999). Additional methods have been published by the EPA, National Institute for Occupational Safety and Health (NIOSH), and others describing analysis procedures for arsenic and some specific arsenic compounds in air (see examples in Table 2). However, most of these published methods do not specify procedures for simultaneously determining multiple individual species of arsenic in air.

Table 2. Examples of sample collection and analysis procedures for arsenic in ambient air

Analysis Method	Method Title	Compounds	Sampling Medium	Analysis
NIOSH Method 7900	Arsenic and compounds as As	All arsenic compounds (as As) exceptAsH3 and As2O3	0.8-μm cellulose ester membrane filter	Flame AA, arsine generation
NIOSH Method 5022	Arsenic, organo	MMA, DMA, arsenic acid	PTFE filter	Ion chromatography/hydride AA
NIOSH Method 6001	Arsine	AsH3	Sorbent tube	GFAA
U.S. EPA	Determination of Arsenic in Suspended PM Collected from Ambient Air	Total As	Glass or quartz fiber filters	GFAA
U.S. EPA	Compendium of Methods for  the Determination of  Inorganic Compounds in  Ambient Air	Total As	Multiple analytical methods	GFAA, ICP, ICP/MS, NAA, etc.
References: NIOSH, 1994a; NIOSH, 1994b; NIOSH, 1994c; EPA, 1988; EPA, 1999.

Within the past two decades, analytical techniques and laboratory instrumentation have been developed and refined to analyze a subset of the individual arsenic species (i.e., As(III), As(V), MMA, DMA, and others) in a variety of environmental matrices, including water, soil, and biological tissues. As an example, Method 1632 was developed by the EPA for analysis of filtered and unfiltered water and tissue using hydride generation and quartz furnace atomic absorption detection (EPA, 2001). More recently, such speciation methodologies have been modified and applied to speciated arsenic in ambient air.

Ambient air can be a very complex matrix, consisting of solid, liquid, and gaseous-phase compounds. The matrix may also be characterized by chemical interferences, for example, as demonstrated by Mukai and Ambe (1987b) In combination with the often very low concentrations of speciated arsenic concentrations in ambient air (on the order of ng/m³), such methods require effective preconcentration and separation techniques.

The remainder of this section summarizes arsenic speciation sampling and analysis techniques developed and performed for the studies identified during the literature review. Although a comprehensive discussion of existing speciated arsenic analytical methods is beyond the scope of this article, the techniques performed will be addressed in a general manner.

Sample collection procedures

Ambient air samples are typically collected for TSP (airborne particles or aerosols less than 100 μm), PM₁₀ (particles with aerodynamic diameter less than 10 μm), or PM_2.5 (particles with aerodynamic diameter less than 2.5 μm) analysis. Samples are often collected using an apparatus that incorporates an inlet, sample collection media, and air-sampling pumps meant to collect air for a specified time period. Sampling inlet size and flow rate generally define the size of the particles to be collected on the sampling medium. Flow rates may be high (30 or more m³/hr), medium (15 or more m³/hr), or low (less than 5 m³/hr) (European Commission, 2000). A high-volume sampling device collects air with a sampling rate held constant over the sampling period (EPA, 1999). For example, when collecting TSP samples, the high-volume design allows the sampled air to be evenly distributed over the surface of a downstream filter, where TSP is collected. The TSP high-volume sampler can be used to determine the average ambient TSP concentration over the sampling period, and, subsequently, the collected material can be analyzed to determine the identity and quantity of inorganic metals present in the TSP (EPA, 1999).

During sample collection, air is drawn through the sampling device and is collected on filter media, such as glass fiber, quartz fiber, membrane filters, and Teflon filters. Generally, the filter medium selected is dependent on the sampling and analysis protocols that will be performed. Due to the generally low concentrations of speciated arsenic in air, it is critical to evaluate filters for background arsenic content, because arsenic may be present at similarly low concentrations in certain filter media. Thus, a “blank” filter must always be analyzed in conjunction with the air samples collected. Many of the studies reviewed reported arsenic content in the associated “blank” filter media, requiring blank correction and subtraction; this was not a unique issue from a temporal standpoint (Johnson and Braman, 1975 Tsopelas et al., 2008).

Numerous techniques were used to collect air samples in the studies reviewed, but they were generally consistent with the standardized procedures defined by U.S. EPA and others (EPA, 1999). For example, Rabano et al. (1989), collected < and >PM_2.5 samples using a high-volume dichotomous virtual impactor (HVDVI) sampler with polytetrafluorene (PTFE) filters. Mukai and Ambe (1987b)and Solomon et al. (1993), also used high-volume air samplers and glass or quartz filters. Rabano et al. (1989), collected samples on PTFE filters for speciated arsenic analyses, and they used a glass microfiber filter for arsenic in TSP. Table 3 summarizes the sampling procedures performed.

Table 3. Ambient air studies and comparison of study methodologies

Study (Reference)	Particle Size	Arsenic (As) Compounds Analyzed	Collection Method	Extraction Method	Analytical Method	QC Reported	Detection Limit
Johnson and Braman, 1975	<3 µm, >3 µm	Inorganic As; methylarsine, dimethylarsine, trimethylarsine	Low vol/glass fiber filter for particles >0.3 µm14 mm pyrex tube/mesh silver plated pyrex beads	Not specified; alkaline wash/extraction	d.c. discharge—emission spectroscopy	Precision: RSD ± 10%	0.1–0.2 ng/m^3
Mukai and Ambe, 1987b	TSP	MMA, DMA, TMA	Hi-vol; quartz fiber filter	1 M HCl/EDTA/6 mL NaBH4	HG-GC-AAS	Accuracy: %R varied from 86–109%	<3–10 ng/m^3 (for all 3 As species)
Mukai and Ambe, 1987a	TSP	MMA, DMA	Hi-vol; quartz fiber filter	NaOH/HCl/NaBH4	HG-GC-AAS	Accuracy: %R = 92%	∼0.1 ng/m^3
Rabano et al., 1989	<2.5 µm, >2.5 µm, TSP	Total As, As(III), As(V)	TSP: Hi-vol/glass microfiber<PM2.5 and >PM2.5: Hi-vol DCVI/PTFE	Total As: HNO3 As(III)/As(V): Ethanol/HCl	Total As method not specified Speciated As: HG-AAS	Accuracy: corr. (r) = 0.94 for Total As versus As(III) + As(V) Accuracy: %R of standards = 95 ± 7/100 ± 8 for As(III)/As(V) in unexposed filters; 79 ± 22/97 ± 23 for As(III)/As(V) in exposed filters Precision: 8% for unexposed filters, 23% for exposed filters	1 ng/m^3
Solomon et al., 1993; Solomon, 1984	TSP	As(III), As(V)	TSP: Hi-vol/quartz fiber filter	Assumed Zn/HCl/NaBH4	Assumed flame-AAS	Precision: CV = 12–13%	0.25 ng/m^3
Talebi, 1999	TSP	Total As, As(III)	Hi-vol/quartz fiber filter	H2SO4/HNO3/water	Flame-AAS	Not reported	Not reported
Šlejkovec et al., 2000	<2 µm, 2–10 µm	Total As, As(III), As(V), MMA, DMA in coarse (2–10 µm) and fine (<2 µm) particles	Stacked filter units	Methanol/water/KH2PO4	Total As-INAAAs(III), As(V), MMA, DMA: HPLC-HG-AFS	RSD <10%	<0.01 ng/m^3 (As(III), MMA)<0.05 ng/m^3 (DMA); <0.08 ng/m^3 (As(V))
Farinha et al., 2004	<2.5 µm, 2.5–10 µm	As(III)/As(V) in PM2.5–10 and <PM2.5	Not specified; described in Reis et al., 1999	Water/CaCl2/H3PO4	Speciated As: HPLC-UV-HG-AFSTotal As: INAA	Total As: 5% analytical error based on single replicate Accuracy: Extraction efficiency = 20.2–78%	Assumed < 0.05–1.1 ng/m^3
Oliveira et al., 2005	TSP	Total As, As(III), As(V), DMA, MMA	Hi-vol/quartz fiber filters	Compared extraction solutions (water, NH2OH·HCl, and H3PO4) and extraction methods (microwave digestion or ultrasonic radiation)	Speciated As: HPLC-HG-AFS Total As: HG-AFS	Optimal mean recoveries [= (As(III) + As(V))/As Total] for the 12 samples were 86% (for the extraction using H3PO4 + microwave digestion) and 94% (for the extraction using NH2OH·HCl + microwave digestion)Stability study indicated no significant change Accuracy: Extraction efficiency Total As versus As(III) + As(V) = 97–126%	Assume <0.2 ng/m^3
Sánchez-Rodas et al., 2007	<10 µm	As(III), As(V) in PM10	Hi-vol/quartz fiber filter	Speciated as: NH2OH·HCl solution/microwave digestionTotal As: HClO4 + HNO3 + HF + digestion	Speciated As: HPLC-HG-AFS Total As: HG-AFS	Accuracy: Extraction efficiency Total As versus As(III)+As(V) = 91-105%Precision and Accuracy: reported to be in range of 3-10% (includes NIST fly ash SRM for Total As) Stability study indicated no significant change (transformation) among As species	0.1 and 0.4 ng/m^3 for As(III) and As(V), respectively
Sánchez de la Campa et al., 2008	<2.5 µm	As(III) and As(V) in PM2.5	Hi-vol/quartz fiber filters	Speciated As: NH2OH·HCl solution/microwave digestionTotal As: HClO4 + HNO3 + HF + digestion	Speciated As: HPLC-HG-AFS Total As: HG-AFS	Accuracy: Total As SRM %R ∼100 Accuracy: Extraction efficiency Total As versus As(III) and As(V) = 91–104% (mean 97%)	0.1 and 0.4 ng/m^3 for As(III) and As(V), respectively
Tsopelas et al., 2008	<2.5 µm, 2.5–10 µm	As(III), As(V), MMA, DMA in PM10 and PM2.5	Hi-vol TSP/glass fiberHi- and low-vol/stacked PFTE Filters: PM10 and PM2.5	HCl	Total/Speciated As: HG-ICP-OES Ion exchange (separation of As(V), MMA, and DMA)	Accuracy: In method development, %R of standards ranged from 94–106%	As(III): 0.1 ng/m^3 As(V): 0.7 ng/m^3 MMA: 0.3 ng/m^3 DMA: 0.4 ng/m^3
Notes: In cases where detection limits are reported with a < symbol, a DL was not reported, but assumed to be < the lowest detected concentration reported in the article. Sample preparation/extraction is final method selected for study/reported results.

Sample extraction procedures

Most arsenic analytical techniques require samples to be extracted or digested in acid prior to instrumental analysis; this approach was used consistently in the majority of the air literature reviewed. Digestion of air sample filter media can occur in open vessels (i.e., in extracted acid on hot plates) or in closed vessels (i.e., microwave digestion in Teflon containers).

Based on our literature review, air samples were prepared in a variety of ways but, in nearly every case, wet digestion or extraction of the air sampling media was performed using an acid (see Table 3). For example, Oliveira et al. (2005) evaluated several extraction methods, comparing water, NH₂OH, HCl, and H₃PO₄ digestion followed by either microwave or ultrasonic radiation; the NH₂OH, HCl, and H₃PO₄ extraction procedures were all determined to be acceptable extraction methods. Sánchez de la Campa et al. (2008), performed sample digestion using HNO₃ and HF, as well as a NH₂OH-HCl digestion method incorporating microwave digestion, a method described by Oliveira et al. (2005), Mukai and Ambe (1987b)also compared several extraction methods, finding that HCl/ethylene diaminetetraacetic acid (EDTA) extraction followed by NaBH₄ reduction was the most efficient, resulting in ∼100% recoveries for both MMA and DMA. Tsopelas et al. (2008), evaluated HCl versus H₃PO₄ extraction of both polycarbonated and glass fiber filters; they determined that HCl led to a higher recovery from both filter types, and the HCl-polycarbonated filter was the optimal preparation method. Thus, although some differences in sample preparation approaches were identified in the articles reviewed, most seemed to be effective for determining speciated arsenic in air samples.

Laboratory analysis methods

Total arsenic air studies generally rely upon standard analytical spectrometry techniques such as atomic absorption spectrometry (AAS) or inductively coupled plasma optical emission spectroscopy (ICP-OES). In AAS, aqueous sample extracts are placed into a graphite tube, evaporated, charred, and atomized (EPA, 2007). During ICP-OES analysis, the sample digestate is nebulized, and the resulting aerosols are transported to a plasma torch for analysis.

The methods described above are sometimes used with regard to arsenic speciation analysis, but they are generally coupled with detectors to facilitate separation of the various species and also improve detection limits. Use of instrumentation such as high-performance liquid chromatography (HPLC; a column chromatography technique used to separate, identify, and quantify compounds based on their polarities and interactions with the column's stationary phase), hydride generation (HG; a technique in which species separation is performed by converting the compound to a volatile hydride), atomic fluorescence spectroscopy (AFS, a technique in which the optical emission from atoms has been excited to higher energy levels by absorption of electromagnetic radiation), ion chromatography (IC; an analytical technique involving separation of ions and polar molecules based on their charge), or mass spectrometry (MS; a technique involving ionization of chemical compounds to generate charged molecules or molecule fragments and measurement of their mass-to-charge ratios) increases analytical sensitivity and allows for good separation of the individual species of arsenic.

Table 3 summarizes the analytical methods performed in the studies reviewed. Methodologies varied among the earliest studies, reflective of a time during which speciation techniques were being developed. For example, Johnson and Braman (1975) used an emission spectroscopy technique to analyze inorganic arsenic (assumed to be As(III) and As(V)), methyl arsine, and dimethyl arsine, and achieved detection limits as low as 0.10 ng/m³; however, the method was not able to distinguish between the arsines and acid forms of arsenic. Solomon et al. (1993), (see also Solomon, 1984) used hydride generation-atomic absorbance spectrometry (HG-AAS) to analyze As(III) and As(V), achieving detection limits as low as 0.20 ng/m³. Mukai and Ambe (1987a 1987b), used HG coupled with gas chromatography-AAS (HG-GC-AAS) to analyze DMA and MMA; the authors achieved detection limits as low as 0.10 ng/m³ for these species. As(III) and As(V) were analyzed by Rabano et al. (1989), using the method developed by Solomon, (1984), HG-AAS; detection limits of 1 ng/m³ were reported in this study. Although detection limits typically decrease as analytical techniques improve, sometimes conditions of the study (such as variations in sample media, preparation techniques, and instrumentation) result in higher detection limits like those reported by Solomon (1984).

Nearly half of the studies used a very sensitive HPLC methodology coupled with hydride generation and atomic fluorescence spectrometry (HPLC-HG-AFS); these studies also have the distinction of being the more recent studies, all performed within the past decade. The technique was able to differentiate between and separate four arsenic species: As(III), As(V), MMA, and DMA. For example, Šlejkovec et al. (2000), analyzed As(III), As(V), DMA, and MMA using this method in coarse and fine PM, achieving detection limits as low as 0.01 ng/m³ for As(III) and MMA, <0.05 ng/m³ for DMA, and 0.08 ng/m³ for As(V). Oliveira et al. (2005), also analyzed As(III), As(V), DMA, and MMA, but in TSP; the detection limits were <0.20 ng/m³ in this study for all species. Studies performed by Sánchez-Rodas et al. (2007), and Sánchez de la Campa et al. (2008), used the method developed by Oliveira et al. (2005), to analyze As(III) and As(V) in PM₁₀ and PM_2.5 samples; detection limits were 0.1 and 0.4 ng/m³, respectively. Farinha et al. (2004), also implemented this technique, with the addition of an ultraviolet (UV) reactor. Unfortunately, actual detection limits were not reported, but detected results were reported as low as 0.0005 ng/m³. Total arsenic was also analyzed during these studies; comparing extraction efficiencies of speciated arsenic (sum of As(III) and As(V)) versus total arsenic generally found efficiencies between 90% and 100% in each of these studies. Tsopelas et al. (2008), also performed speciated arsenic analyses using an HG method but instead coupled the technique with an ICP detector, achieving detection limits as low as 0.10 ng/m³ for As(III), 0.70 for As(V), 0.30 ng/m³ for MMA, and 0.40 ng/m³ for DMA. The HG-ICP analyses demonstrated excellent accuracy and precision as well.

There are several potential limitations for speciated arsenic analyses. For example, in HG analysis, arsenic forms a volatile hydride (arsine, AsH₃) when reduced. Within a certain pH range, arsine is produced only by the trivalent forms of arsenic, and not the pentavalent forms. In some HG techniques, As(III) is determined under controlled pH conditions, and As(V) is determined by subtracting As(III) from the total arsenic results. It should be noted that DMA and MMA can form volatile hydrides during analysis of some matrices; thus, caution should be used if performing this speciation technique when these species may be present. As an example, Šlejkovec et al. (2000), noted that some forms of arsenic species can sometimes undergo alteration during sampling or sequential extraction procedures; for instance, they noted that As(III) can undergo approximately 20% oxidation, which would underestimate As(III) content and overestimate As(V) content. Farinha et al. (2004), described this possibility as well, stating that As(V) may have been oxidized from As(III) during sampling, storage, and extraction. This potential issue was evaluated in the Tsopelas et al. (2008), study using HCl as an extraction acid, although a different analytical methodology was used (HG-ICP versus the HPLC-HG-AFS technique used by Šlejkovec et al. (2000), no issues regarding oxidation were noted. From the studies reviewed, it appears that HG methodology, coupled with various detectors, is the method of choice, particularly as it has proven to be highly sensitive for separation of As(III) and As(V).

Quality assurance and quality control (accuracy, precision, comparability, and representatives)

Measurement systems can be characterized by systematic and random errors that have the potential to undermine the quality of data results. Thus, studies are designed to minimize such errors, incorporating QA/QC procedures to promote overall study data reliability and usability. Good QA/QC procedures help to ensure that the results generated are accurate, precise, comparable, and representative. This section will explore some of these procedures, discuss QA/QC results described in the literature reviewed, and consider how the results may or may not impact the data reported. Table 3 presents a summary of the QC results reported in the studies reviewed.

Accuracy

Accuracy is a measure of the bias in a measurement system that may result from sampling or analytical error, and it is an important element in determining overall data reliability. Sources of error that may contribute to poor accuracy are sampling inconsistency, laboratory contamination, and matrix interferences. Accuracy may be evaluated through several different QC parameters, including via laboratory control sample (LCS) or standard reference material (SRM) recoveries, matrix spike recoveries, and extraction efficiencies.

Rabano et al. (1989), evaluated accuracy in several ways during their study. First, they evaluated recovery of speciated arsenic standards on unexposed PFTE filters; recoveries were on the order of 95% and 100% for As(III) and As(V), respectively. This was followed by determination of recoveries of known concentrations of speciated arsenic standards added to exposed filters (similar to matrix spikes); recoveries of the standards were 79% ± 22% and 97% ± 23% for As(III) and As(V), respectively. The authors concluded that the complex nature of the PM slightly affected measurement of the arsenic species; matrix interferences may also have been an issue but, in general, recoveries were very good.

Several of the studies reviewed evaluated accuracy in terms of extraction efficiency, comparing the summed concentrations of speciated arsenic (i.e., As(III) + As(V)) to the total arsenic results and calculating % recoveries (see Table 3). As an example, to evaluate the efficiency of their extraction procedures, Sánchez de la Campa et al. (2008), compared the results of PM_2.5 to total arsenic in dry-ashed samples; comparing the sum of the As(III) and As(V), extraction efficiencies ranged from 91%–41%, with a mean of 97%, indicating excellent recovery. Most of the other studies reported similar efficiencies, indicating that the methods were in control (see Table 3).

Stability studies offer another potential method of evaluating study accuracy. This type of study evaluates a species' chemical stability at select points during the overall analytical process, such as through sample handling, storage, preparation, and analysis. Based on our review, little information is available on the stability of arsenic species in air, but two studies did evaluate this endpoint. Sánchez-Rodas et al. (2007), evaluated short- (within 1 day of sampling) and long- (sample analyzed monthly for 1 year) term chemical stability in a sample exhibiting elevated arsenic concentrations; no increase, decrease, or conversion of arsenic species (As(III) and As(V)) was noted. Oliveira et al. (2005), also evaluated sample stability for both total and speciated arsenic (As(III) and As(V)) after both extraction and speciation analyses. According to their findings, no changes in the distribution of arsenic species occurred during a 33-month period. Neither of these studies reported how the samples were stored during the stability analysis. Therefore, at least among these studies, the form of arsenic species appeared stable over time in the sample for a given method.

As an analyte, arsenic is prone to matrix interferences, which can affect accuracy by suppressing or enhancing the sample matrix, resulting in false negatives or positives. Many published analytical methods described the various interferences that can occur, including spectral, physical, or chemical interferences. Chemical interferences are rather uncommon and are usually characterized by molecular compound formation, ionization effects, or solute vaporization effects. Spectral interferences are more likely during arsenic determinations and may occur during either AAS or ICP analyses. For example, during AAS analysis, arsenic can be affected by light scattering and nonspecific absorption (EPA, 2007). Background correction is often employed to prevent false positives or biased high results. During ICP-MS analysis, isobaric interferences may occur that require background corrections; the use of internal standards may also help eliminate the interferences (EPA, 1994). During HG-AAS analysis, elevated concentrations of certain metals (chromium, cobalt, copper, mercury, molybdenum, nickel, or silver) can cause interferences due to generation of arsine. Mukai and Ambe (1987b) aware that various elements (such as iron, calcium, and aluminum) could interfere with the reduction of methylarsenic compounds, performed an in-depth analysis to evaluate potential interferences from different metal mixtures on MMA and DMA. The authors noted that a mixture of iron and nitrate produced strong interferences on quantitation of DMA. Nitrite was also shown to interfere quite strongly. To address these potential interferences in their own speciation analyses, Mukai and Ambe (1987b) added EDTA to the samples during the extraction process. Ultimately, this method resulted in recoveries of methylarsenic compounds on the order of 100%, indicating control of interferences and demonstrating excellent accuracy.

Several studies incorporated the use of SRMs, which are certified standards of known concentration carried through the entire preparation and analytical process. In all cases, the SRMs evaluated total, not speciated, arsenic. Nonetheless, where reported, the SRMs indicated excellent recovery, indicating good method accuracy for total arsenic. For example, Tsopelas et al. (2008), reported recoveries of 98–103%; Sanchez de la Campa et al. (2008), reported nearly 100% recovery. Although certified reference materials are available for total arsenic and were evaluated in several studies (see Sánchez-Rodas et al., 2007), limited information is available on the identity and concentrations of specific arsenic species in some SRMs (Agency for Toxic Substances and Disease Registry, 2007).

In summary, most of the studies reviewed indicated good accuracy, an indication that the analytical methods performed were in control.

Precision

Precision is defined as a measure of the reproducibility of individual measurements of the same property under a given set of conditions. It is a qualitative measure of the variability of a group of data compared to its average value. Precision is generally monitored through the use of duplicate analyses, with results expressed in terms of relative percent difference (RPD) or relative standard deviation (RSD).

In general, the studies reviewed reported good precision, although this QC parameter was not reported consistently across studies. For example, Johnson and Braman (1975) reported analytical precision of ±10%, whereas Rabano et al. (1989), reported precision error of 8% for unexposed filters versus 23% for exposure filters, based on the addition of speciated arsenic standards to these matrices. Šlejkovec et al. (2000), reported a study RSD of <10%.

Comparability

Comparability is a somewhat qualitative parameter that expresses the confidence with which data sets can be compared. Comparable data allow for the ability to combine analytical results acquired from various sources using different methods for samples collected over a specified time period or collected from different studies. Comparability relies upon precision and accuracy within the individual data sets to be acceptable to ensure confidence in the data sets. The use of consistent analytical and sampling methods is essential in ensuring that separate data sets are comparable. In addition, comparability is often affected by QA/QC criteria, such as sample preservation, holding times, blank contamination, method detection/contract-required quantitation limits, and calibrations.

Although there are many studies describing total arsenic in ambient air, few published studies have evaluated a full range of arsenic species in air. As described, even the studies reviewed that did make this distinction focused primarily on As(III) and As(V). The major issues were with regard to comparability across studies (i.e., different species were analyzed, and in most cases, different sampling, preparation, and analysis methods were performed). In addition, it remains unclear how representative the samples' analyses were of average ambient conditions, as most studies represent conditions across a month or two. Also, as nearly all studies were conducted in urban/industrial locations, there remains a relative paucity of sampling in natural background conditions occurring outside the influence of anthropogenic emission sources. All of these factors make it difficult to compare the various techniques to determine an optimal approach. In addition, the variation in sample collection, preparation, and analysis methods also limit identification of the most robust method(s). Comparability between studies remains significantly hindered, making it difficult to draw any conclusions, particularly when considering the early studies. That said, the more recent studies performed by Oliveira et al. (2005), Sánchez-Rodas et al. (2007), and Sánchez de la Campa et al. (2008), appear more comparable due to the use of similar extraction and analysis methods, indicating that the HG-HPCL-AFS method is reproducible with high precision and accuracy.

Representativeness

Representativeness is the degree to which a single measurement is indicative of the characteristics of a larger sample or area. More specifically, it is the degree to which the data gathered accurately and precisely represent the actual field conditions—in this case, ambient air.

In most studies, samples were collected over time periods ranging from several weeks up to several years with variable intervals between samples. The studies performed in Japan by Mukai and Ambe (1987a 1987b) represent spring and early summer conditions in 1987. Šlejkovec et al. (2000), reported results from May 1998. Farinha et al. (2004), and Tsopelas et al., 2008 both sampled over multiple years; 1994–1995 and 2004–2006, respectively. The Spanish studies also span several years; Oliveira et al. (2005), sampled monthly in 2000, whereas Sánchez-Rodas et al. (2007), and Sánchez de la Campa et al. (2008), sampled every other week in 2001 and 2002. The California studies (Rabano et al., 1989; Solomon et al., 1993) collected samples in January 1987 and April–May 1989. Studies with short durations provide a snapshot of conditions that exist at the time of sampling, whereas longer, multiyear studies are more representative of average conditions over time.

Trends

The articles reviewed encompass a time frame of more than 30 years (1975–2008), during which significant improvements were made in analytical instrumentation and detection limits. Although sensitive detection limits were possible in 1975 (the publication date of the first article reviewed), over time, sample preparation procedures, species separation techniques, and analytical detection methods were refined and improved to provide more effective chemical species analysis. In addition, a few studies (e. g., Mukai and Ambe, 1987b Oliveira et al., 2005) compared multiple extraction methods in an attempt to identify the most efficient technique and to allow selection of the most robust methods. Based on the studies reviewed, extraction of quartz fiber filters in NH₂OH·HCl or HCl appeared to be the optimal preparation method, whereas HG-HPLC-AFS or HG-AAS seemed to be preferred analytical methods. The procedures described by Tsopelas et al. (2008), also seemed to be very effective in analyzing the four arsenic species.

In general, instrumentation, detection limits (from 1 ng/m³ to below 0.25 ng/m³), and species separation techniques have greatly improved over the past 30 years. Nonetheless, although there is some consistency in terms of analytical instrumentation used across studies, the sample collection/preparation techniques varied. In addition, the specific species reported differed across studies. Thus, trend analysis and data comparison are limited, ultimately making it difficult to draw conclusions regarding the best analysis techniques; in general, however, method accuracy and precision were in control.

Given the few articles identified in the literature, and the variety of sampling and analysis procedures implemented, we would recommend that future studies employ consistent sample preparation and analysis methodologies for ease of data comparison across studies. In addition, consistent use and reporting of QC samples (e.g., blanks, SRMs) would be helpful for data evaluation and interpretation.

Arsenic Toxicology

An understanding of both speciated arsenic exposure and toxicity are necessary for evaluating the human health implications of arsenic in air. The toxicity of inorganic arsenic is well studied, particularly via the oral route. Due to high concentrations of inorganic arsenic in groundwater that exist in many parts of the world (e.g., Taiwan (Chen et al., 1992), Chile, (Marshall et al., 2007), Bangladesh (Chen et al., 2009)), many studies have investigated the relationship between consumption of inorganic arsenic in drinking water and health effects in humans. Overall, these studies have confirmed an association between ingested inorganic arsenic and lung, bladder, and skin cancer (Agency for Toxic Substances and Disease Registry, 2007; National Research Council, 2001). Inorganic arsenic-associated noncancer effects have also been observed, including skin lesions and cardiovascular effects. (Agency for Toxic Substances and Disease Registry, 2007; National Research Council, 2001).

Inhalation studies on arsenic exposure are mainly in the form of occupational studies. Most of the available information comes from copper smelter studies, where worker exposure to arsenic occurred as a mixture of arsenic trioxide in dust and vapor (the relative amounts of exposure via vapor versus dust are not quantified, but, depending on job function, exposure to both fumes and dust can occur). (Enterline and Marsh, 1982 Lee-Feldstein, 1983; Lee-Feldstein, 1986;Pinto et al., 1977). In these studies, arsenic was usually measured as total arsenic in air or urine with no evaluation on the exact form of inhaled arsenic (some studies did speciate arsenic in urine, but, even if this analysis took place, the form of inhaled arsenic is not possible to determine). Overall, these occupational studies have found a consistent relationship between arsenic and lung cancer. Noncancer effects, such as respiratory irritation, nausea, skin effects, and neurological effects, have also been reported but with less frequency or consistency than lung cancer (Agency for Toxic Substances and Disease Registry, 2007; European Commission, 2000).

The dependency of arsenic toxicity on valence state has been known for many decades, but information mainly comes from oral studies or in vitro studies, and many of the studies involve short-term exposures. The discussion below summarizes some of the available scientific research on the differential toxicity among the different arsenic compounds. Although oral and in vitro studies may have uncertain relevance for assessing potential risk from inhalation exposures, and particularly chronic exposure, we include this discussion (which begins with a discussion of in vitro and oral animal studies and graduates to a discussion on inhalation-based information) to highlight some of the toxicological differences in arsenic compounds and potential implications for inhalation exposure. Additionally, by understanding what information on speculated arsenic toxicity does exist, we are able to bring greater attention to further research needs with respect to inhalation toxicology of speciated arsenic.

Based on oral studies in animals (as well as in vitro studies), it is generally recognized that As(III) is more toxic than As(V) in both acute and chronic scenarios—although there are very few in vivo studies evaluating the overt toxicity of both compounds in a single animal experiment (Agency for Toxic Substances and Disease Registry, 2007). ATSDR (2007), reports that, overall, As(III) is 2–3 times more toxic than As(V). Many in vitro studies have tested the toxicity of As(III) and As(V) in a common test system and have observed that As(III) is significantly more toxic than As(V). As an example, Cohen et al. (2002) observed that the LC₅₀ for As(III) (sodium arsenite) in human bladder cells was 4.8 µM, whereas the LC₅₀ for As(V) (sodium arsenate) was 31.3 µM. Similarly, using the lactate dehydrogenase (LDH) assay, As(III) (LC₅₀ = 68 ± 16.9 µM) was considerably more cytotoxic than arsenate (LC₅₀ = 1,628 ± 110 µM) in a cell culture of human hepatocytes (Petrick et al., 2000). An in vitro study specifically investigating the comparative toxicity of arsenic trioxide and arsenic pentoxide (the arsenic species typically found adsorbed to PM in air) found that arsenic trioxide was about 4–5 times more toxic than arsenic pentoxide, depending on cell type (Tse et al., 2008).

The toxicity of arsenic is also dependent on methylation status, largely because methylated arsenic compounds undergo limited cellular uptake and are mainly excreted without metabolism to arsenic metabolites that are more toxic (see discussion of trivalent arsenic metabolites below) (Cohen et al., 2006). Substantial scientific evidence exists indicating that oral exposures to pentavalent methylated arsenic compounds (e.g., MMAV and DMAV) are far less toxic than exposure to inorganic arsenic compounds (Cohen et al., 2006). For example, the LD₅₀s for As(III) in rats range from 15 to 145 mg As/kg, whereas the rat LD₅₀ for MMA is 2833 mg As/kg (male and female combined), and the rat LD₅₀ for DMA is 1935 mg/kg·day (Agency for Toxic Substances and Disease, 2007; EPA, 2006). The lower toxicity of the methylated arsenic compounds via the oral route is observed with intermediate and chronic exposures as well (Agency for Toxic Substances and Disease Registry, 2007), and it has been shown repeatedly in in vitro studies (Cohen et al., 2002; Petrick et al., 2000). The pentavalent methylated metabolites of arsenic are generated from the metabolism of inorganic arsenic in animals (including humans), but they also occur naturally in the environment and serve as the active ingredient in the pesticides monosodium methanearsonate (MSMA), disodium methanearsonate (DSMA), calcium acid methanearsonate (CAMA), and cacodylic acid (U.S. Environmental Protection Agency, 2006).

Unlike the pentavalent methylated arsenic compounds, which have low toxicity compared to inorganic arsenic compounds, research over the past decade has demonstrated that methylated arsenic compounds in the trivalent state (e.g., monomethylarsonous acid [MMA(III)] and dimethylarsinous acid DMA(III)]) are associated with potent cytotoxicity and genotoxicity (Cohen et al., 2006). These compounds, which are unstable in the environment, are generated in vivo from the metabolism of inorganic arsenic and are hypothesized to play a key role the carcinogenic properties of inorganic arsenic. Recent research has also suggested that arsenic is metabolized to a thiol-containing intermediate that may be involved in arsenic carcinogenicity via the oral route (Suzuki et al., 2010).

As noted above, studies of the relationship between toxicity and the valence and methylation state of arsenic have mainly focused on oral exposures, and there has been much less study of arsenic toxicity from exposure via the inhalation and dermal routes. Although the metabolism of inhaled inorganic arsenic is not well studied and may vary as a function of dose and particle size, it is noteworthy that studies of arsenic metabolism show that, after absorption, inorganic arsenic compounds are extensively methylated and excreted in the urine in similar proportions regardless of the exposure pathway (i.e., the relative proportion of inorganic arsenic, MMA, and DMA are similar via the inhalation and oral pathway) (Morton and Mason, 2006 Vahter, 1999;Yager et al., 1997). Also, the limited information available on differential inhalation toxicity of the different forms of inorganic arsenic suggests patterns similar to oral toxicity. For example, acute toxicology studies show that pentavalent methylated arsenic compounds are less toxic than inorganic arsenic compounds (As(III) and As(V)). Specifically, a developmental study in rats demonstrated that a 1-day exposure of about 76 mg/m³ of arsenic (in the form of arsenic trioxide) resulted in the death of all pregnant dams. In contrast, the LC₅₀ for DMA was 3900 mg/m³ in a 2-hr exposure in rats. Also, a sodium salt of MMA (DSMA) was associated with very low acute toxicity. In rats, exposure to 6100 mg DSMA/m³ was not associated with any deaths (Stevens et al., 1979).

Research on the comparative toxicity (acute or chronic) of As(III) versus As(V) via inhalation is limited. In one study, Lantz et al. (1994), showed that, compared to As(V), As(III) was a more potent inhibitor of some indicators of pulmonary macrophage function in vitro, but this distinction was not as clear in vivo (after intratracheal instillation in rats). In another experiment, Marafante and Vahter (1987), showed that only 0.06% of As(III) and 0.02% of As(V) was retained in the lung 3 days after intratracheal instillation in hamsters, suggesting that both forms of arsenic in the lungs can be fully absorbed. Therefore, it is reasonable to hypothesize that As(III) and As(V) would exhibit the same difference in toxicity seen with oral exposures, which are also almost completely absorbed into the blood stream, although the reduction capacity of lung cells—particularly following inhalation exposures—is largely unstudied. However, as intratracheal instillation studies are performed using a preparation of As(III) and As(V) in water, information from these studies may not be completely applicable to arsenic bound to PM due to likely differences in bioavailability and absorption kinetics (Yager et al., 1997). Variability in the absorption and toxicity of different forms of arsenic-bound PM has not been evaluated. Also, these studies are short-term studies; chronic exposure studies of speciated arsenic via inhalation are lacking in the literature.

Although the dependency of valence state and methylation status on overall toxic severity may be similar between the oral and inhalation route, toxic endpoints from inorganic arsenic exposure do differ by exposure route, suggesting that yet unknown factors, perhaps related to metabolism, may influence route-specific target organs. For example, whereas oral exposure to inorganic arsenic is associated with bladder, lung, and skin cancers, inhalation of inorganic arsenic is only associated with lung cancer. For noncancer endpoints, there is some, but not complete, overlap of target organ toxicity. For example, there is some evidence (albeit limited for the inhalation route) that both inhalation and oral exposures can cause skin effects and peripheral neuropathy, but liver effects are only seen with oral arsenic exposure. Also, the link between arsenic exposure and cardiovascular disease is much better established for oral exposures. As mentioned above, this suggests some qualitative or quantitative differences in arsenic metabolism and/or disposition between the oral and inhalation pathways, but specific differences have not yet been fully elucidated. Understanding these differences, particularly in light of the fact that excretion products are similar between the two exposure routes, is a key research need.

Implications of Risk Assessment and Research Needs

Although there is widespread recognition that arsenic toxicity is dependent on form, arsenic risks have been traditionally evaluated by examining the relationship between total arsenic and health effects. This has been largely based on the assumption that environmental forms of arsenic are mainly in the more toxic inorganic form. Although this may be an appropriate assumption under certain circumstances, overall, the research investigating arsenic speciation in air, from both an exposure and a toxicology standpoint, has been lacking. In this paper, we have attempted to perform a comprehensive review of the available information on arsenic speciation in air. Understanding the current state of knowledge allows for a critical evaluation of data gaps and an opportunity to prioritize arsenic inhalation risk assessment needs.

It should be noted that although arsenic speciation, as well as the speciation of other metals, will refine risk evaluations, most air toxics monitoring programs in the United States do not routinely report information on speciated metals at monitoring sites. One exception is EPA's National Air Toxics Trends Station (NATTS) network, which requires monitoring for hexavalent chromium, the most toxic form of chromium. Similarly, U.S. EPA air monitoring programs and air toxics risk assessments for metals are currently focused on the PM₁₀ fraction, whereas the National PM_2.5 Speciation Program reports metals concentrations in PM_2.5 across U.S. sites (EPA, 2011). Thus, even when the scientific information on arsenic speciation in air with regard to particle size is more complete, there are additional challenges in translating this improved understanding into more practical risk applications.

Exposure considerations

Our analysis of the reviewed literature provides further confirmation that most of the arsenic present in air is in the inorganic form (As(III) or As(V)), whereas the concentrations of organic arsenic compounds (MMAV and DMAV) are mainly undetectable. This is consistent with summary literature that has also reported that the inorganic form predominates in air. For example, EPA (EPA, 1984) reports that arsenic in ambient air is usually a mixture of particulate As(III) and As(V), and that, in general, the organic species are of negligible importance except where methylated arsenic pesticide application or biotic activity have occurred. Although our analysis is in agreement regarding inorganic arsenic, we did not find significant concentrations of methylated arsenic compounds, even in areas with arsenical pesticide use; the basis for EPA's conclusion regarding the organic arsenic compounds is not apparent.

ATSDR (2007) also reports that air mainly contains arsenic in the inorganic form, but cites only the 1984 EPA document as a reference. A 2000 report by the European Commission (EC) (2000) relied on a subset of the studies that are reviewed here to similarly conclude that air mainly contains inorganic form, with As(V) predominating.

Although our review is consistent with inorganic arsenic as the predominating form of arsenic in air, it is interesting to note that inorganic arsenic, overall, accounted for about 90% of the total arsenic measured, even in cases where the methylated compounds were specifically measured. It is unclear whether an undefined arsenic fraction was somehow not measured, or whether there were analytical limitations that did not fully account for the As(III) and As(V) present in the samples. Other forms of arsenic can exist in air; for example, arsenic trisulfide has been reported from coal combustion, organic arsines from oil combustion, and arsenic trichloride from refuse incineration (EPA, 1982). The extent to which the analytical methods reviewed in this paper were able to account for these or other arsenic fractions—and whether they would be meaningful contributors to ambient air and, importantly, risk—remains uncertain.

In addition to arsenic form, particle size is another important factor when considering human exposure and potential risk. Different particulate fractions are associated with different health risks. TSP, which reflects any particle that is small enough to travel via air, is unlikely to be the best measure of potential human exposure because only a fraction of TSP (i.e., <PM₁₀) is respirable. Although both PM₁₀ and PM_2.5 can enter and accumulate in the lungs, PM_2.5 may be more toxicologically active because it can penetrate further into the lungs and be absorbed into the bloodstream.

Although information is limited, research suggests that total arsenic concentrations are concentrated in the smaller size fractions (i.e., PM_2.5). We also observed some evidence of this trend in the reviewed studies, but results were inconsistent. Rabano et al. (1989), found arsenic enrichment in fine particles compared to coarse particles. In this study, As(III) predominated in the fraction larger than 2.5 µm, whereas As(V) was slightly more prevalent when the sample was restricted to PM_2.5. Likewise, Tsopelas et al., (2008) who measured speciated arsenic levels in an industrial area of Greece, found that total arsenic concentrations in PM_2.5 and PM_10–2.5 were 1.9 and 1.1 ng/m³, respectively; however, TSP had the highest total arsenic concentration at 3.4 ng/m³. All arsenic was in the form of As(V) across all size fractions. Concentrations of arsenic by particle size varied by source in the study by Šlejkovec et al. (2000), In samples collected in a tunnel, arsenic concentrations in particles between 2 and 10 µm were higher than in particles smaller than 2 µm. The opposite pattern was observed in an open city square. This study also found that the arsenic in the finer fraction was more soluble, which may affect bioavailability. Given the variability in results, it is clear that more research in this area is warranted, specifically with regard to how As(III):As(V) ratios might vary with particle size and source.

It is well established in the literature (and consistent with the research presented here) that total exposures to inorganic arsenic in air—particularly compared to exposures from food and water—contribute a negligible amount to total arsenic exposure. In the present evaluation, total arsenic air concentrations, which mainly represent urban or industrial environments, ranged from 0.96 to 18.9 ng/m³. This is consistent with estimates by EC (2000), which reports that concentrations of arsenic in air range from 0 to 1 ng/m⊃ in remote areas, 0.2 to 1.5 ng/m⊃ in rural areas, 0.5 to 3 ng/m⊃ in urban areas, and up to about 50 ng/m⊃ in the vicinity of industrial sites. Likewise, the 2002 annual median arsenic concentration in PM_2.5 for remote areas across the United States was 0.18 ng/m³ (McCarthy and Hafrer, 2006). Based on these data, EC (2000), estimated that, in relation to food, cigarette smoking, water, and soil, air contributes less than 1% of total arsenic exposure—even when assuming an arsenic air exposure that is significantly above typical background (20 ng/m³). A likely reason that arsenic risk assessment research has been directed to the study of other environmental media relates to this general recognition that air is a minor arsenic exposure pathway for the typical individual. As noted earlier, occupational exposure to inhaled arsenic, which can involve very high concentrations—historically upwards of 3 mg/m³—of arsenic has received the most study and remains an important issue in arsenic inhalation risk assessment.

Toxicity considerations

As described earlier, information on the relative toxicity of arsenic species via inhalation is limited. At present, there is a paucity of information on the comparative toxicity of As(III) and As(V) via inhalation. Most information must be extrapolated from oral studies. And although it is likely that toxicity data from oral studies may provide some useful information on inhalation toxicity due to similarities in metabolism, at present, oral toxicity studies are deficient in fully elucidating the differences in toxicity between As(III) and As(V). In particular, differences in the metabolizing capacity of lung cells (via inhalation exposures) versus the metabolizing capacity of liver cells via oral exposures may provide insight into route-specific similarities and differences in toxicity. Additionally, although it is generally assumed that As(III) is about 2–3-fold more toxic than As(V) (Agency for Toxic Substances and Disease Registry, 2007), direct animal evidence to support this quantitative relationship, particularly for inhalation exposures, is limited.

Most investigation into any potential differences of toxicity between As(V) and As(III) via inhalation has occurred in experiments using intratracheal instillations (Lantz et al., 1994). Although this approach may be useful for establishing relative differences in potencies between As(V) and As(III), it is not as relevant for understanding potential human risk implications. Human exposure to arsenic in air mainly occurs through association of arsenic-associated PM, which has differences in bioavailability and toxicokinetics. We were not able to locate any studies that evaluated the toxicity of different forms of arsenic associated with PM. The lack of toxicity of speciated forms of arsenic, particularly in the context of exposure from different PM fractions, represents a key data gap in arsenic inhalation risk assessment.

Information on the relative toxicity of methylated arsenic compounds and inorganic arsenic compounds via inhalation is also limited. However, because of the negligible concentrations of methylated arsenic compounds in air and their overall low toxicity, airborne methylated arsenic compounds are not expected to pose a significant human risk.

When conducting risk assessments in the United States, risk assessors preferentially rely on toxicity criteria published in the Integrated Risk Information System (IRIS). The inhalation toxicity criterion in IRIS, which is used to predict the carcinogenic risk from arsenic via inhalation, is based on studies in workers exposed to arsenic trioxide in both vapor and dust. Because the arsenic in these studies was almost entirely in the trivalent form and not always associated with PM, it is possible that risks posed by occupational exposures may be different than those posed by environmental exposures. Arsenic in environmental exposures would be expected to be a mixture of As(III) and As(V) (with more As(V), as noted in Table 1) and largely associated with PM, two features that are likely to be associated with lower toxicity. From this perspective, traditional arsenic risk assessment may overestimate arsenic risk from ambient air, although more investigation would be necessary to quantify the magnitude of the impact. This line of research is particularly important, because a recent paper has suggested that background concentrations of total arsenic in PM_2.5 can present a cancer risk greater than 1 × 10⁻⁶ (McCarthy et al., 2009).

Recommendations and Conclusions

Compared to other environmental media, the study of speciated arsenic in air has been limited. Although we conducted a comprehensive review of measures of speciated arsenic in air, the analytical methods, point sources of arsenic, and PM fractions measured were too variable to draw any meaningful conclusions regarding how arsenic composition may vary by point source or the potential risk implications. To facilitate more meaningful comparisons across studies, it is important to harmonize sample preparation and analytical techniques. Studies should incorporate and document QA/QC measures such that a thorough evaluation of study accuracy, precision, representativeness, and comparability is made possible. Also, a consistent method of collection and analysis is needed. Accurate measurement of air pollutants is critical, particularly if assessment of human health impacts will result in regulatory actions by federal, state, and local governments.

Finally, although the study of toxic air pollutants has been the subject of interest and concern for many years, it is clear from our review that little work regarding arsenic speciation in air, particularly in the United States, has been recorded in the literature. The limited research conducted to date, however, has consistently shown that arsenic in air is mainly in the form of inorganic arsenic (As(III) + As(V)). Concentrations of methylated arsenic compounds (MMA and DMA) are very low (in the pg/m³ range), even in areas expected to have the highest concentrations. Because of low concentrations of the methylated compounds in air and their overall low toxicity, research directed at further understanding the toxicity of MMA and DMA via air is unlikely to yield information that would further improve our ability to estimate arsenic inhalation risks. Investigation into the differences in the inhalation toxicity of As(III) versus As(V), in conjunction with exposure assessments aimed at understanding how As(III):As(V) ratios change by source and particle size, stands to offer the most significant insights into arsenic inhalation risk assessment.

Acknowledgments

This work was sponsored by the Electric Power Research Institute. The authors would like to thank Sagar Thakali and Gautham Jegadeesan for their assistance in performing background research; Jasmine Lai and Bethany Allen for their help in editing and preparing the manuscript; and Barbara Beck for providing her scientific input.