Arsenic Toxicology and Cancer Risk

RoseyFields evaluates the toxicology of Arsenic using the following evidence base: the U.S. EPA’s 2025 IRIS Toxicological Review of Inorganic Arsenic (final report), the Agency for Toxic Substances and Disease Registry’s Medical Management Guideline for Arsenic and Inorganic Arsenic Compounds, and the peer-reviewed primary literature identified within and supplementing those sources. Because the 2025 IRIS assessment concentrates its formal evidence synthesis on human epidemiological dose-response relationships for cancer and noncancer outcomes, the animal toxicology, mechanistic, and comparative sections below draw substantially on the primary experimental studies cited in that review and identified through a supplemental review of the peer-reviewed toxicology literature.

1. Chemical Identity and Physicochemical Properties

Arsenic (As; CAS No. 7440-38-2) is a naturally occurring metalloid with chemical and physical properties intermediate between a metal and a nonmetal. It exists in four oxidation states (−3, 0, +3, and +5) and is rarely found in elemental (zero-valent) form in the environment because of its reactivity. In biological and environmental systems, arsenic occurs predominantly as inorganic arsenite (As[III]) and arsenate (As[V]) oxyanions, as well as a range of methylated organic species. Elemental (grey, metallic) arsenic is a steel-grey solid (density 5.7 g/cm3; melting point 817°C at 28 atm; boiling point 613°C, sublimes) that is intermediate in reactivity between metals and nonmetals. Arsenic trioxide (As2O3), the most toxicologically significant and prevalent inorganic form, is an odorless, tasteless, white-to-transparent solid (molecular weight 197.8; melting point 312°C; density 3.7 g/cm3) with appreciable water solubility (approximately 17 g/L at 16°C), which contributes to its ready absorption following ingestion.

Inorganic arsenic compounds generally predominate over organic arsenicals in air, water, and soil, and inorganic species are substantially more toxic than the organoarsenicals found in seafood (e.g., arsenobetaine, arsenocholine, and arsenosugars), which are considered essentially nontoxic at typical dietary exposure levels. Arsenic is not appreciably flammable in solid inorganic forms but can generate highly toxic arsine gas (AsH3) upon contact with strong acids, reducing metals, or nascent hydrogen, and arsenic trioxide fumes are released when arsenic-bearing materials are burned at high temperature.

With respect to environmental persistence and mobility, inorganic arsenic does not degrade; it cycles among oxidation states and chemical species through microbially- and abiotically-mediated redox and methylation reactions but remains in the environment indefinitely as arsenic. Arsenic is generally immobile to moderately mobile in soils, with mobility increasing under reducing conditions and in sandy, low-clay soils, and adsorption enhanced by iron, aluminum, and manganese oxide content (Welch et al., 1988). In water, arsenic speciation and transport depend on oxidation-reduction potential and the presence of competing sorbents; soluble pentavalent arsenate predominates and adsorbs to iron oxides, aluminum hydroxides, manganese compounds, and clays, while more soluble and mobile trivalent arsenite becomes more prevalent under reducing (e.g., anoxic groundwater) conditions. Atmospheric arsenic is associated with fine particulates (mass median diameter approximately 1 μm) generated principally by high-temperature anthropogenic processes such as coal and oil combustion, nonferrous metal smelting, and waste incineration, and these particles can be transported over distances of thousands of kilometers before wet or dry deposition (Pacyna, 1987). Arsenic does not bioaccumulate substantially up the food chain in the manner of persistent lipophilic organics; body burdens in mammals reflect ongoing exposure and are cleared with a urinary half-life on the order of days rather than accumulating over a lifetime, although arsenic can be retained for longer periods in keratinized tissues (hair, nails, skin) and, to a lesser degree, in bone.

2. Toxicokinetics (ADME)

2.1 Absorption

Inorganic arsenic is efficiently absorbed following oral exposure; ingested soluble arsenite and arsenate are absorbed from the gastrointestinal tract to a high degree in both experimental animals and humans, which is the toxicokinetic basis for the primacy of the oral route in both the acute toxidrome and the chronic disease evidence base. Dermal absorption of solid or particulate inorganic arsenic (e.g., arsenic trioxide dust) is comparatively minor under intact skin conditions, though direct contact can produce local irritation. Inhalation of arsenic-bearing particulates results in absorption from the respiratory tract that is generally insufficient, for arsenic trioxide specifically, to produce acute systemic toxicity at typical incidental exposure levels, although occupational and combustion-related exposures to other, more bioavailable inorganic arsenic compounds (e.g., lead arsenate, arsenic sulfide) can represent a more significant inhalation hazard.

2.2 Distribution and Metabolism

Following absorption, inorganic arsenic is distributed throughout the body and undergoes a complex, interdependent sequence of enzymatic and nonenzymatic oxidation-reduction and methylation reactions. The generally accepted biotransformation pathway (Figure summarized in Sams et al., 2007) involves reduction of pentavalent arsenate to trivalent arsenite, followed by oxidative methylation — catalyzed principally by arsenic (+3) methyltransferase (AS3MT), using S-adenosylmethionine (SAM) as the methyl donor — to monomethylarsonic acid (MMA) and dimethylarsinic acid (DMA), with further minor conversion to trimethylated species. Both pentavalent and the more reactive, more toxic trivalent methylated intermediates (MMA[III], DMA[III]) are generated during this process. Methylation is generally viewed as a detoxification pathway that facilitates urinary excretion, although the trivalent methylated intermediates are, on a molar basis, more potent inhibitors of cellular enzymes and more cytotoxic and genotoxic than inorganic arsenite itself in many in vitro systems, so biomethylation does not uniformly equate to detoxification at the cellular level.

A central toxicokinetic finding, with direct relevance to the interpretation of animal data, is that there are substantial interspecies differences in arsenic metabolism. Using cultured primary hepatocytes from multiple species, Drobná et al. (2010) demonstrated that the rate and extent of arsenic methylation, and the proportion of trivalent versus pentavalent methylated metabolites produced and retained, differ markedly across species; some commonly used laboratory species (e.g., certain rodent strains, marmosets, chimpanzees) methylate inorganic arsenic far less efficiently than humans, while others (e.g., mice, rats) methylate it more efficiently. These interspecies differences complicate the selection of animal models that faithfully recapitulate human internal dosimetry and have historically impeded the development of rodent bioassay models for arsenic carcinogenesis, since rodents both metabolize and, in the case of rats, sequester arsenic in erythrocytes very differently than humans do.

In mice specifically, Kenyon et al. (2008) characterized tissue distribution and urinary excretion of inorganic arsenic and its methylated metabolites in C57BL/6 mice following subchronic oral exposure to arsenate in drinking water, finding dose-dependent accumulation in kidney, liver, and other tissues along with efficient urinary elimination predominantly as DMA, consistent with rodents’ comparatively efficient methylating capacity relative to less efficient methylating species. Snow et al. (2005b) similarly characterized tissue dosimetry in the mouse following chronic dietary or drinking-water exposure to several arsenic compounds, providing internal dose data used in physiologically based pharmacokinetic (PBPK) model development and evaluation.

Human genetic variation in the AS3MT gene and related methylation pathway genes has also been shown to influence individual arsenic methylation efficiency and, in turn, susceptibility to arsenic-associated disease. Wood et al. (2006) characterized functional variation in human AS3MT through gene resequencing and functional genomics, identifying polymorphisms with reduced methyltransferase activity. Pierce et al. (2012) conducted a genome-wide association study in a Bangladeshi population identifying variants in the chromosome 10q24.32 region encompassing AS3MT associated with arsenic metabolism efficiency, and Engström et al. (2013) similarly linked the AS3MT haplotype to DNA methylation and expression changes at multiple genes surrounding the AS3MT locus, providing a mechanistic basis for population-level variability in urinary metabolite profiles and in susceptibility to arsenic toxicity.

2.3 Excretion and Physiologically Based Pharmacokinetic (PBPK) Modeling

Arsenic and its methylated metabolites are excreted predominantly in urine, with a relatively short blood elimination half-life (on the order of hours), such that blood arsenic concentrations can return to background within a day or two of a single exposure even though urinary excretion of the absorbed dose continues over a longer period; a normal blood value therefore does not exclude a clinically significant recent exposure. Several PBPK models for inorganic arsenic have been developed to relate external oral exposure to internal dose and urinary biomarker concentrations (El-Masri and Kenyon, 2008; El-Masri et al., 2018). The El-Masri-Kenyon model, evaluated against two large human biomonitoring datasets from Bangladesh and Nevada comprising matched drinking-water and urinary arsenic measurements, establishes an approximately 1:1 empirical relationship between daily inorganic arsenic intake and total urinary arsenic excretion across a wide dose range, and this model has been used to support dose-reconstruction from urinary biomarker data in the epidemiological literature (Kenyon, 2021).

3. Mechanism(s) of Action / Mode of Action

Arsenic toxicity and carcinogenicity are not attributable to a single unifying mechanism; rather, EPA’s mode-of action analysis and the broader mechanistic literature converge on several interrelated pathways that appear to operate across multiple target organs and across both cancer and noncancer endpoints. These include: (1) generation of reactive oxygen species (ROS) and induction of oxidative stress; (2) high-affinity binding of trivalent arsenicals to vicinal dithiol groups on proteins, inhibiting key enzymes (e.g., pyruvate dehydrogenase and other mitochondrial dehydrogenase complexes); (3) structural mimicry of phosphate by pentavalent arsenate, permitting substitution into high-energy phosphate intermediates and uncoupling of oxidative phosphorylation; (4) impairment of cell-cycle control and DNA damage-repair processes; (5) epigenetic dysregulation, including altered DNA methylation and histone modification; (6) endocrine-disrupting activity; and (7) cytotoxicity with compensatory regenerative cell proliferation.

Of these, oxidative stress has the broadest evidentiary support as a mode of action operative across cardiovascular disease, diabetes, hepatic injury, lung and bladder carcinogenesis, neurotoxicity, nonmalignant respiratory disease, adverse pregnancy outcomes, renal injury, and dermal effects (Kitchin and Conolly, 2010; Jomova et al., 2011; Flora, 2011). Arsenic metabolism itself, through cycling between trivalent and pentavalent oxidation states, consumes cellular thiol pools (notably glutathione), generates trivalent methylated intermediates, and produces superoxide, hydrogen peroxide, and hydroxyl radicals, thereby perturbing cellular redox balance and depleting antioxidant defenses. Oxidative stress biomarkers have been measured across a range of in vitro systems at low micromolar arsenic concentrations and in animal studies at oral doses in the range of approximately 0.5–1.7 mg/kg day.

Downstream of ROS generation, arsenic exposure has been shown to alter expression of antioxidant-response proteins, including molecular changes associated with cell transformation observed by Li et al. (2011) in human keratinocyte models, and to perturb DNA-repair enzyme activity, including DNA polymerase β (Snow et al., 2005a) and the base-excision repair enzyme OGG1 and persistence of DNA damage following chronic monomethylarsonous acid exposure (Wnek et al., 2009). Persistent, low-level activation of the Nrf2-Keap1 antioxidant pathway, rather than a transient adaptive response, has been proposed as a mechanism by which chronic arsenic exposure promotes a pro-survival, pro-proliferative cellular phenotype resembling the constitutive Nrf2 activation observed in several human tumor types (Lau et al., 2013). In human urothelial cell models, arsenic and its trivalent monomethylated metabolite (MMA[III]) have been shown to activate mitogenic RAS/ErbB2 signaling (Eblin et al., 2007), induce ROS-dependent transformation-associated changes via modulation of MAPK signaling (Eblin et al., 2009), stimulate interleukin-8 over-production and autocrine cell activation (Escudero-Lourdes et al., 2012), and damage DNA via AP-1 activation (Drobna et al., 2003). Kojima et al. (2009), using As3mt knockout mice with impaired arsenic biomethylation capacity, demonstrated that biomethylation of inorganic arsenic is required for arsenic-associated oxidative DNA damage in vivo, directly linking the metabolic pathway described in Section 2 to the genotoxic mode of action. In the vasculature, Straub et al. (2008) showed in mice that arsenic stimulated hepatic sinusoidal capillarization required NADPH oxidase-generated superoxide, and Soucy et al. (2003, 2005) demonstrated that chronic arsenic exposure stimulates angiogenesis and angiogenic gene expression in vivo through ROS-dependent signaling, mechanistically linking oxidative stress to arsenic’s vascular and hepatic effects.

4. Animal Toxicological Evidence by Organ System

The following subsections summarize representative experimental animal studies evaluating the toxicity of inorganic arsenic (principally as sodium arsenite or sodium arsenate administered in drinking water or by gavage) across major organ systems. Study designs, species, dose levels, and principal findings are noted for each.

4.1 Hepatic

The liver is a major target of arsenic toxicity in rodents, consistent with its role as the principal site of arsenic biomethylation. Xie et al. (2004) administered sodium arsenite, sodium arsenate, MMA(V), or DMA(V) in drinking water to v-Ha-ras transgenic (Tg.AC) mice at doses of 150, 200, 1,500, and 1,000 ppm as arsenic, respectively, for 17 weeks; all arsenical forms produced dose-dependent hepatic arsenic accumulation, reduced body-weight gain, and mortality in some treatment groups, with arsenic accumulation in liver occurring largely independent of the administered chemical species. Straub et al. (2008), in a mechanistic mouse study, found that chronic arsenic exposure induced hepatic sinusoidal capillarization (a structural precursor lesion associated with impaired hepatic microcirculation and portal hypertension) via an NADPH oxidase/superoxide-dependent pathway, providing a structural and mechanistic basis for arsenic-associated liver injury observed in longer-term rodent bioassays. Interspecies hepatocyte metabolism studies (Drobná et al., 2010) further indicate that the liver’s biotransformation capacity, and therefore its exposure to the more reactive trivalent methylated intermediates, differs substantially by species, which is relevant when extrapolating rodent hepatic findings to humans.

4.2 Immune System

Nain and Smits (2012) conducted an 18-week subchronic drinking-water study in juvenile male Wistar rats at arsenite concentrations of 0, 0.4, 4, and 40 ppm, examining clinical biochemistry, B-cell-mediated and innate immune responses, and gross and histopathology alongside oxidative-damage biomarkers (lipid, protein, and nucleic acid oxidation). Immunological and pathological indices were identified as sensitive markers of subclinical arsenic toxicity, detectable at exposure levels below those producing overt clinical signs, supporting the immune system as a relatively sensitive noncancer target of chronic low-level arsenic exposure in rodents.

4.3 Cardiovascular

Multiple rodent studies have examined arsenic’s cardiovascular and vascular effects using drinking-water exposure paradigms. In apolipoprotein E-deficient (ApoE−/−) mice, a genetic model of accelerated atherosclerosis, Srivastava et al. (2007) demonstrated that in utero arsenic exposure induced early onset of atherosclerotic lesion formation in offspring, and Srivastava et al. (2009) subsequently showed that postnatal arsenic exposure exacerbated atherosclerotic lesion formation and vascular inflammation in the same model. Soucy et al. (2003) reported that arsenic stimulated angiogenesis and tumor growth in vivo in a mouse xenograft model, and Soucy et al. (2005) further characterized dose-related neovascularization and upregulation of angiogenic gene expression in mouse tissues following chronic arsenic exposure, implicating dysregulated angiogenic signaling in arsenic-associated vascular pathology. In rats, subchronic (90-day) drinking-water exposure to sodium arsenite (25, 50, and 100 ppm) produced dose-dependent oxidative stress in the thoracic aorta (increased lipid peroxidation and hydrogen peroxide, decreased glutathione and antioxidant enzyme activities), together with dyslipidemia (reduced HDL-cholesterol, increased triglycerides, total cholesterol, and LDL-cholesterol) and electrolyte disturbances (Waghe et al., 2014; Sarath et al./Waghe et al., 2017), supporting oxidative and lipid-mediated mechanisms underlying arsenic-associated vascular dysfunction observed in epidemiological studies of ischemic heart disease.

4.4 Renal

Renal effects have been documented as part of broader multi-organ subchronic toxicity studies; the Nain and Smits (2012) rat study noted above included renal biochemical and histopathological endpoints among its subclinical toxicity indicators, and subchronic rat drinking-water studies evaluating arsenic-induced vascular oxidative stress (Waghe et al., 2014; Sarath et al./Waghe et al., 2017) also reported dose-related increases in relative kidney weight and altered renal biochemical indices (e.g., electrolyte handling) alongside hepatic changes, consistent with the kidney’s central role in arsenic and arsenic-metabolite excretion and its resulting exposure to high local concentrations of arsenic and its methylated species.

4.5 Reproductive

Male reproductive toxicity has been demonstrated in multiple rat studies. Huang et al. (2016) conducted an integrated proteomic and metabolomic analysis of rat testis following arsenic exposure, finding decreased serum testosterone and reduced sperm quality (motility and count) accompanying broad alterations in testicular protein and metabolite profiles related to spermatogenesis, sperm function, and fertility, with the pattern of changes suggesting mediation through ERK/AKT/NF-κB-dependent signaling. Additional rodent studies have reported arsenic-induced reductions in sperm motility, viability, and daily sperm production together with disruption of the hypothalamic pituitary-gonadal axis (reduced testosterone, follicle-stimulating hormone, and luteinizing hormone) and testicular histopathological changes (Leydig cell and seminiferous tubule injury), with oxidative stress markers (elevated reactive oxygen species and thiobarbituric acid-reactive substances, reduced antioxidant enzyme activity) implicated as a proximate mechanism.

4.6 Endocrine

Beyond the hypothalamic-pituitary-gonadal axis effects noted above, arsenic has been characterized in the mechanistic literature as an endocrine-disrupting agent capable of interfering with steroidogenic enzyme activity and glucocorticoid and thyroid hormone signaling pathways in rodent models; adrenal and thyroid tissue have been identified as target organs in longer-term rodent bioassay work (see also Section 5 for adrenal tumor findings in the mouse transplacental carcinogenesis studies), consistent with cross-organ endocrine involvement in arsenic’s noncancer and cancer toxicity profile.

4.7 Neurological / Developmental Neurotoxicity

A substantial body of rodent evidence links developmental and perinatal arsenic exposure to neurobehavioral and neurochemical deficits. Zhao et al. (2017) exposed pregnant mice to 0, 25, 50, or 100 mg/L arsenite in drinking water from gestation through postnatal day 35 and evaluated hippocampal synaptic structure and protein expression (PSD-95, synaptophysin) in offspring at multiple postnatal timepoints; spatial learning was impaired at the lowest tested dose (25 mg/L), spatial memory was impaired at the highest dose (100 mg/L), and postsynaptic density thickness and synaptic protein expression were reduced in a dose-related manner, providing a structural correlate for arsenic-associated learning and memory deficits. In rats, perinatal (gestation day 6 through postnatal day 21) oral arsenic exposure (2 or 4 mg/kg-day) produced hypoactivity and impaired spatial learning and memory in offspring, associated with decreased striatal dopamine D2 receptor expression, reduced tyrosine hydroxylase protein, and altered striatal dopamine levels, implicating dysfunction of the dopaminergic system in arsenic developmental neurotoxicity. Zhu et al. (2017) further demonstrated that combined fluoride and arsenic exposure in drinking water impaired spatial memory in rat offspring and activated the ERK/CREB signaling pathway in hippocampus and cortex, while additional rodent studies have reported arsenic-associated alterations in glutamatergic signaling (NMDA and metabotropic glutamate receptor expression) and reduced basal hippocampal synaptic transmission following low-dose (as low as 50 ppb) as well as high-dose gestational and early-life arsenic exposure, together indicating that multiple neurotransmitter systems and synaptic structural elements are sensitive targets of developmental arsenic exposure across a wide dose range.

5. Animal Carcinogenicity Bioassay Findings

Historically, development of a rodent model of inorganic arsenic carcinogenesis proved difficult using conventional adult-exposure, single-agent drinking-water bioassay designs, in part reflecting the interspecies metabolic differences discussed in Section 2. This impasse was substantially resolved through a series of transplacental (in utero) and whole-life exposure studies conducted primarily by Waalkes and colleagues at the National Institute of Environmental Health Sciences, which established inorganic arsenic as a complete, multi-site transplacental carcinogen in mice.

In the foundational study, Waalkes et al. (2003) administered sodium arsenite in drinking water (0, 42.5, or 85 ppm as arsenic) to pregnant C3H mice from gestation day 8 through day 18 and observed offspring for up to two years; prenatal arsenic exposure produced dose-related increases in hepatocellular carcinoma and adrenal cortical adenoma in male offspring, and dose-related increases in ovarian tumors and lung carcinoma in female offspring. Waalkes et al. (2004) reproduced and extended these findings, confirming inorganic arsenic as a transplacental carcinogen in mice and characterizing the promotional effects of postnatal 12-O-tetradecanoylphorbol-13-acetate (TPA) exposure, which enhanced hepatic and pulmonary but not dermal tumor yields. Using a second mouse strain (CD1) and postnatal hormonal-modulator co-exposures, Waalkes et al. (2006a, cited in the secondary literature) reported that in utero arsenic exposure induced urogenital tract tumors (including uterine tumors) in female offspring, exacerbated by postnatal diethylstilbestrol treatment, and Waalkes et al. (2006b, cited in the secondary literature) reported enhanced urinary bladder and liver carcinogenesis in male CD1 mice following transplacental arsenic exposure combined with postnatal diethylstilbestrol or tamoxifen. Waalkes et al. (2007) reviewed and further replicated this body of work, concluding that brief in utero arsenic exposure in mice consistently induces tumors or preneoplastic lesions in liver, lung, urinary bladder, adrenal, kidney, ovary, uterus, oviduct, and vagina of adult offspring across three temporally separate studies and two mouse strains (C3H and CD1), with liver, lung, urinary bladder, and kidney findings concordant with recognized human target sites of arsenic carcinogenicity. Notably, despite skin being a major human target organ for arsenic-induced cancer, these transplacental mouse studies did not produce dermal tumors, suggesting either a tissue-specific limitation of the transplacental mouse model or a requirement for additional co-carcinogenic stimulation (e.g., UV or phorbol ester promotion) to elicit arsenic-associated skin carcinogenesis in this species.

Tokar et al. (2011) extended this line of investigation to whole-life, human-relevant low-dose exposure, administering 0, 50, 500, or 5,000 ppb (μg/L) arsenic as sodium arsenite in drinking water to breeder CD1 mice beginning three weeks before mating and continuing through gestation, lactation, and up to two years of direct offspring exposure post-weaning. This study found that whole-life exposure at doses spanning the low parts-per billion range relevant to human drinking-water exposure was carcinogenic in the offspring, with tumor induction that did not require the higher parts-per-million maternal exposures used in the earlier transplacental-only study designs, reinforcing the biological plausibility of arsenic carcinogenicity at exposure levels closer to those experienced by human populations.

Taken together, this series of mouse bioassays (Waalkes et al., 2003, 2004, 2007; Tokar et al., 2011) constitutes the principal experimental animal carcinogenicity evidence base for inorganic arsenic, demonstrating multi-site tumor induction — including at organs concordant with recognized human arsenic target sites (liver, lung, bladder, kidney) — predominantly following developmental (in utero and early postnatal) rather than adult-only exposure, a pattern that has informed subsequent discussion of early-life susceptibility windows in the human epidemiological and risk assessment literature.

6. Comparison of Animal-Derived Findings to Human Epidemiological Evidence

The overall hazard conclusions from the rodent carcinogenicity bioassay literature are broadly concordant with the human epidemiological evidence: both lines of evidence implicate the liver, lung, kidney, and urinary bladder as arsenic carcinogenicity target organs, and both lines of evidence identify early-life (prenatal and early postnatal) exposure as producing effects, including cancer, that may not be apparent, or may only be fully expressed, later in life. However, there are important points of divergence and limitation that affect how the animal data should be weighed alongside human evidence in a weight-of-evidence assessment.

First, skin cancer — one of the most well-established human arsenic cancer endpoints — has not been reproduced in the mouse transplacental carcinogenicity models, and inorganic arsenic administered alone, without additional promotional or co-carcinogenic stimulation, generally fails to produce skin tumors in rodent skin-carcinogenesis models generally. This discordance likely reflects toxicokinetic and toxicodynamic differences between rodent and human skin and systemic arsenic handling rather than an absence of biological plausibility, since arsenic’s oxidative and genotoxic mode of action is not inherently tissue-restricted; however, it means the rodent database alone would not have flagged skin as a target organ, underscoring a documented limitation of relying on animal bioassay data in isolation for this chemical.

Second, the well-documented interspecies differences in arsenic biomethylation efficiency (Drobná et al., 2010) mean that internal dose and the relative proportion of reactive trivalent methylated metabolites reaching target tissues can differ substantially between commonly used rodent strains and humans at a given external (administered or ingested) dose. This is a principal reason that conventional adult-only rodent bioassays historically failed to reproduce human-relevant arsenic carcinogenicity, and it is also the reason PBPK models calibrated to human biomonitoring data (El-Masri and Kenyon, 2008; El-Masri et al., 2018) rather than simple administered-dose extrapolation from rodent bioassays have been preferred as the basis for human dose-response and risk assessment in the current IRIS evidence synthesis, which relies primarily on human epidemiological dose-response data (e.g., for bladder cancer, lung cancer, ischemic heart disease, and diabetes) rather than on animal bioassay-derived toxicity values.

Third, several noncancer organ-system effects well documented in rodent studies — including cardiovascular oxidative stress and dyslipidemia (Waghe et al., 2014; Srivastava et al., 2007, 2009), immune perturbation (Nain and Smits, 2012), reproductive/endocrine effects (Huang et al., 2016), and developmental neurotoxicity (Zhao et al., 2017) — are qualitatively consistent with, and mechanistically supportive of, human epidemiological associations between arsenic exposure and cardiovascular disease, immune dysfunction, adverse reproductive outcomes, and neurodevelopmental deficits in children. In these instances the animal data serve primarily a mechanistic and biological-plausibility function, corroborating and helping to explain associations observed in epidemiological studies, rather than serving as the primary quantitative basis for dose-response assessment, which for arsenic is instead grounded in the extensive human data available from highly exposed populations (e.g., Bangladesh, Taiwan, Chile, and portions of the United States with elevated groundwater arsenic).

Finally, the magnitude and pattern of exposure used in many rodent noncancer toxicity studies (e.g., 25–150 ppm sodium arsenite in drinking water, corresponding to internal doses well above typical human environmental exposure) is considerably higher than typical human environmental or even moderately elevated groundwater exposure levels, which more often fall in the tens to low hundreds of micrograms per liter range. While such higher dose rodent studies are valuable for characterizing mechanism and identifying target organs, direct quantitative extrapolation of rodent no-observed-adverse-effect levels or benchmark doses to human risk requires the toxicokinetic adjustments described above and is generally considered less reliable, for arsenic specifically, than dose-response modeling grounded directly in human epidemiological data, which is the approach taken in EPA’s current IRIS assessment.

7. Data Gaps and Areas of Ongoing Research

Several data gaps and areas of continuing scientific uncertainty remain relevant to the animal and mechanistic toxicology of inorganic arsenic:

• Absence of a validated rodent model for arsenic-induced skin carcinogenesis following inorganic arsenic exposure alone, which limits the ability of animal bioassay data to corroborate this well-established human target organ and leaves open questions about the tissue-specific determinants of arsenic skin carcinogenicity in humans.
• Incomplete characterization of the relative contribution of specific arsenic metabolites (inorganic arsenite/arsenate versus trivalent versus pentavalent methylated species) to specific target-organ toxicities, particularly for noncancer endpoints such as cardiovascular disease and diabetes, where the mechanistic sequence of molecular initiating events and key events remains only partially elucidated relative to the cancer mode-of-action work.
• Persistent uncertainty regarding the most appropriate animal species and strain for extrapolating internal dosimetry to humans, given well-documented and substantial interspecies (and, within species, individual genetic) variability in arsenic methylation capacity; continued refinement and cross-validation of PBPK models against expanding human biomonitoring datasets remains an active area of work.
• Limited direct animal dose-response data at exposure levels most relevant to typical U.S. environmental and drinking-water exposure (single- to low double-digit micrograms per liter), since most rodent mechanistic and organ-toxicity studies to date have used considerably higher drinking-water concentrations; the Tokar et al. (2011) whole-life low-dose mouse carcinogenicity study is a notable exception, but comparable low-dose rodent data for noncancer endpoints (cardiovascular, immune, endocrine, reproductive, neurodevelopmental) remain comparatively sparse.
• Incomplete mechanistic understanding of sex-specific and life-stage-specific susceptibility differences observed in both animal and human studies, including the biological basis for the pronounced susceptibility of the prenatal and early postnatal period to arsenic carcinogenesis and neurodevelopmental toxicity relative to adult-only exposure.
• Continued need for research clarifying the quantitative relationship between oxidative stress biomarkers measured in animal and in vitro systems and apical adverse outcomes in humans, to strengthen the quantitative linkage between the mode-of-action evidence summarized in Section 3 and human dose-response relationships.

References