Ethylene Oxide (EtO) Toxicology and Cancer Risk

RoseyFields evaluates the toxicology of Ethylene oxide (EtO) using the following evidence base.

1. Chemical Identity and Physicochemical Properties Relevant to Toxicity

Ethylene oxide (oxirane; CAS No. 75-21-8) is the simplest cyclic ether, a three-membered ring epoxide with the formula C2H4O and a molecular weight of 44.05 g/mol (WHO 2003). At room temperature it exists solely as a colorless, flammable gas with a sweet, ether-like odor; its boiling point is 10.6°C and its vapor pressure at 20°C is extremely high (1,095 mm Hg), so any liquid or dissolved-phase EtO volatilizes rapidly (NIOSH 2016; WHO 2003). The odor threshold in air (787 mg/m3, roughly 433 ppm) lies far above concentrations that produce systemic toxicity, so smell provides no reliable warning of hazardous exposure (Amoore and Hautala 1983).

The properties that govern EtO’s environmental behavior and internal disposition all trace to the strained epoxide ring, a highly electrophilic, reactive moiety. EtO is completely miscible with water and most or ganic solvents (log Kow approximately -0.22 to -0.30), indicating a polar, non-lipophilic character (Hansch and Leo 1979; WHO 2003). This low partition coefficient predicts, and empirical review confirms, that EtO does not bioaccumulate in animal or plant tissue; rapid metabolic clearance via hydrolysis and glutathione conjugation, combined with the compound’s polarity, precludes meaningful bioconcentration in the food chain (WHO 1985). EtO is also non-persistent in environmental media relative to many industrial chemicals, though estimates vary by matrix. In air, the dominant degradation route is oxidation via hydroxyl-radical attack, with modeled half-lives ranging from roughly 38 to 382 days (Fritz et al. 1982; WHO 2003), while di rect reaction with molecular oxygen is negligibly slow, on the order of centuries (Bogan and Hand 1978). In water, EtO undergoes both volatilization (half-life on the order of 1 hour) and hydrolysis to ethylene glycol, with a hydrolytic half-life of approximately 9–14 days depending on temperature, pH, and salinity (Conway et al. 1983; Long and Pritchard 1956); reaction with hydroxyl radicals in water is far slower, with a com puted half-life near 50 years (Anbar and Neta 1967). A multimedia fugacity model estimated atmospheric persistence localized to the discharge area of only a few days once dilution and removal are considered to gether (WHO 2003), and a separate multimedia fate model found predicted environmental concentrations posed low hazard quotients for terrestrial and aquatic species (Staples and Gulledge 2006). Because EtO does not sorb appreciably to soil or sediment, environmental persistence is governed primarily by rapid abi otic loss rather than accumulation, which shapes exposure assessment toward inhalation as the dominant human exposure pathway, particularly in occupational settings such as chemical production and medical device/food sterilization.

The same electrophilic reactivity that limits environmental persistence is directly responsible for EtO’s toxicological potency: as an alkylating agent, EtO reacts spontaneously (without enzymatic activation) with nucleophilic sites in proteins, RNA, and DNA, a property discussed further under Mechanism of Action be low.

2. Toxicokinetics: Absorption, Distribution, Metabolism, and Excretion

Absorption. EtO is efficiently absorbed across the respiratory epithelium. In hospital sterilization work ers exposed to workplace air concentrations of 0.11–13.2 ppm, approximately 75% of inhaled EtO was ab sorbed into the bloodstream, with blood:air partition coefficients of 2.5–3.3 measured from venous samples (Brugnone et al. 1985, 1986). Animal inhalation studies in rats, mice, and rabbits confirm rapid pulmonary uptake, with absorption approaching 100% in mice exposed briefly at low concentrations (Ehrenberg et al. 1974; Filser and Bolt 1984; Tardif et al. 1987). No human or animal data address oral absorption, and the only dermal absorption estimate—a permeation rate of 0.125 mg/(cm2·hour) through excised human skin exposed to a 1% aqueous solution—suggests that concentrated liquid or dissolved EtO can also cross intact skin (cited via WHO/IARC compilations).

Distribution. No human distribution data exist, but rodent studies show that absorbed EtO distributes rapidly and broadly. Following nose-only inhalation exposure, peak tissue concentrations in rats and mice were comparable across blood, muscle, and brain, with testicular concentrations running 20–50% lower than blood (Brown et al. 1996). Blood elimination half-times were markedly shorter in mice (2.4–3.1 min utes) than in rats (10.8–13.8 minutes), a species difference with implications for internal dose comparisons discussed below (Brown et al. 1996). In vitro-derived tissue:air partition coefficients were broadly similar across fat, brain, lung, liver, blood, and testes in rats, arguing against strong tissue-specific sequestration (Krishnan et al. 1992). Radiolabel studies found the highest EtO-derived activity in lung, liver, kidney, and urinary bladder, with comparatively low levels in fat (Ehrenberg et al. 1974; Tyler and McKelvey 1982).

Metabolism. EtO is detoxified through two competing pathways: (1) hydrolysis (enzymatic and spontaneous) to ethylene glycol, with further oxidation to oxalic and formic acid and ultimately CO2, and (2) conjugation with glutathione, catalyzed by glutathione-S-transferases, yielding mercapturic acid and thiodiacetic acid metabolites (Scheick et al. 1997). The relative contribution of each pathway is markedly species-dependent. Physiologically based pharmacokinetic (PBPK) modeling estimated that glutathione conjugation accounts for approximately 80% of metabolized EtO in rats, 60% in mice, and only about 20% in humans, with the hydrolytic pathway making up the remainder in inverse proportion (Fennell and Brown 2001). Despite this difference in pathway partitioning, simulated peak blood concentrations were relatively similar across the three species after accounting for differences in uptake and metabolic capacity, with area under-the-curve values ranking humans > rats > mice (Fennell and Brown 2001). This species divergence in metabolic routing is toxicologically important: because humans rely more heavily on hydrolysis and less on the glutathione-dependent pathway than rodents, tissue glutathione depletion—a phenomenon docu mented in mice at higher exposure concentrations and correlated with sublinear (saturating) increases in blood EtO levels (Brown et al. 1998)—may be less protective in extrapolating rodent internal dosimetry di rectly to humans. Sex-related differences in metabolism appear minimal based on available data (Fennell and Brown 2001; Mertes et al. 1985), whereas age-related differences may be clinically relevant: because the detoxification enzymes involved (hydrolases and glutathione-S-transferases) are incompletely developed in early life, young children could exhibit reduced metabolic clearance and, consequently, increased internal dose per unit external exposure relative to adults (McCarver and Hines 2002; Zhong et al. 2018). Genetic polymorphism further modulates individual susceptibility: hemoglobin adduct levels were roughly two-fold higher in workers with a null GSTT1 genotype (reduced glutathione-conjugation capacity) than in those with a functional genotype (Yong et al. 2001), and urinary excretion of the glutathione-derived con jugate was correspondingly elevated in workers with a functional (non-null) GSTT1 genotype (Haufroid et al. 2007).

Excretion. EtO and its metabolites are cleared rapidly, predominantly via the kidney. In rats exposed by in halation to 14C-labeled EtO, 59% of recovered radioactivity appeared in urine, 12% as exhaled 14CO2, 4.5% in feces, and about 1% as unmetabolized EtO in expired air (Tyler and McKelvey 1982). In mice, an average of 78% of absorbed radioactivity was excreted in urine within 48 hours of a brief exposure (Ehrenberg et al. 1974), and EtO elimination kinetics were first-order under closed-system inhalation conditions (Filser and Bolt 1984). Urinary metabolites identified in exposed rats include ethylene glycol, 2-hydroxyethyl mercapturic acid, and thiodiacetic acid (Scheick et al. 1997).

A complicating feature of EtO toxicokinetics—relevant to interpreting both animal and human biomonitor ing data—is that EtO is also produced endogenously, via lipid peroxidation, methionine and heme oxidation, and gut microbial metabolism of ethylene (Thier and Bolt 2000). Because hemoglobin and DNA adducts cannot easily distinguish exogenous from endogenous EtO, background adduct levels are non-zero even in unexposed individuals; modeling of hemoglobin N-(2-hydroxyethyl)valine (HEV) adduct data suggests that background (non-occupational, non-smoking) adduct levels correspond to an equivalent continuous air ex posure on the order of 1–2 ppb (Kirman and Hays 2017), a consideration that must be factored into both risk assessment and epidemiological exposure reconstruction.

3. Mechanism(s) of Action

EtO’s principal, and most thoroughly documented, mechanism of toxicity and carcinogenicity is direct chem ical alkylation. As a strained three-membered epoxide ring, EtO reacts non-enzymatically with nucleophilic centers in macromolecules, forming stable hydroxyethyl adducts with proteins (including hemoglobin), RNA, and DNA. The predominant DNA lesion is N7-(2-hydroxyethyl)guanine (7-HEG), with lesser amounts of N3-(2-hydroxyethyl)adenine and O6-(2-hydroxyethyl)guanine also detected (Walker et al. 1990; Walker et al. 1992b). Following inhalation exposure of rats, 7-HEG adducts accumulated in a duration-related manner across brain, lung, spleen, kidney, liver, and testis (Walker et al. 1990), and comparable DNA adduct forma tion has been documented following co-exposure to ethylene, consistent with a shared alkylating pathway for exogenous and endogenously-derived EtO (Rusyn et al. 2005).

This direct DNA-reactive mechanism underlies the consensus classification of EtO as a mutagenic carcinogen. EtO consistently induces gene mutation, chromosomal aberrations, sister chromatid exchange, micronu cleus formation, DNA strand breaks, and cell transformation across bacterial, fungal, insect, mammalian, and human test systems in vitro, and produces gene mutation, chromosomal aberrations, sister chromatid exchange, micronucleus formation, dominant lethal mutation, and heritable translocations in vivo in ro dents and, to a variable extent, in occupationally exposed humans. As IARC summarized, “ethylene oxide consistently acts as a mutagen and clastogen at all phylogenetic levels” and induces dose-related cytoge netic damage in the lymphocytes of exposed workers. This body of genotoxicity evidence, combined with the demonstrated formation of DNA adducts at target organs for carcinogenesis, supports a mutagenic mode of action in which EtO-induced heritable DNA damage in somatic (and germline) cells is considered a key initiating event for tumor development, rather than a secondary consequence of cytotoxicity or other indirect mechanisms. Because this mechanism does not require metabolic activation, and because EtO re acts directly and stochastically with DNA at any exposure level above zero, regulatory bodies have generally not identified a toxicological threshold for the mutagenic/carcinogenic mode of action, in contrast to the pattern typically assumed for many non-genotoxic noncancer endpoints.

For noncancer endpoints, mechanistic understanding is less complete. For peripheral neuropathy, a pro posed mechanism involves impaired axonal transport: in rats exposed repeatedly to EtO, anterograde axonal transport velocity in peripheral nerve was reduced by roughly one-third relative to controls, ac companied by increased degeneration of myelinated fibers, suggesting that disruption of axonal transport machinery—potentially via alkylation of transport-associated proteins—contributes causally to the periph eral neuropathy seen in both animals and highly exposed workers (Nagata et al. 1992). Inhibition of creatine kinase activity has also been proposed as contributing to EtO-induced neuromuscular effects (Matsuoka et al. 1990). For hematological and hepatic effects, mechanistic data are comparatively sparse, and the pro file’s authors note that findings for several noncancer endpoints (e.g., adrenal and hematologic changes in a key chronic rat bioassay) are confounded by concurrent bacterial infection in the study colony, complicating mechanistic interpretation.

4. Animal Toxicological Evidence by Organ System

Nearly all animal toxicity data for EtO derive from inhalation studies; oral data are limited to a small number of gavage studies, and dermal/ocular data are limited largely to irritation and sensitization testing.

Respiratory. Acute high-concentration exposure produces classic irritant injury: dyspnea, pulmonary edema, pulmonary hemorrhage, and congestion were reported in rats, mice, and guinea pigs exposed for brief periods at 357–841 ppm (Hollingsworth et al. 1956; NTP 1987), with rhinitis in mice at 400 ppm for up to two weeks (NTP 1987). Intermediate-duration exposures produced labored breathing and nasal discharge in rats at 406 ppm for six weeks and pulmonary congestion in dogs at 292 ppm for six weeks (Jacobson et al. 1956), and increased relative lung weight in rats and guinea pigs at 113–204 ppm for up to 226 days (Hollingsworth et al. 1956). In the pivotal two-year rat bioassay, bronchopneumonia and suppurative rhinitis occurred at 50 ppm, but all groups—including controls—experienced a concurrent bacterial infection, confounding attribution to EtO (Lynch et al. 1984a, 1984b). No respiratory effects were seen in a two-generation reproduction study at 100 ppm (EPA 1994) or in mice exposed for 102 weeks at up to 100 ppm (NTP 1987).

Cardiovascular. Data are limited. No electrocardiographic abnormalities were found in monkeys exposed repeatedly at 50 or 100 ppm for up to two years, and no histopathological cardiovascular lesions occurred in chronically exposed mice (Lynch et al. 1984b; NTP 1987).

Gastrointestinal. No histopathological GI effects were seen in rats or mice exposed by inhalation at 100 ppm for up to 104 weeks (Lynch et al. 1984a, 1984b; NTP 1987). By the oral route, gastric irritation occurred in female rats gavaged at 100 mg/kg/day, with a no-observed-adverse-effect level (NOAEL) of 30 mg/kg/day (Hollingsworth et al. 1956).

Hematological. Decreases in hemoglobin, hematocrit, and erythrocyte count, with reticulocytosis, were reported in rats, mice, and dogs exposed at 250–500 ppm for 6–13 weeks (Fujishiro et al. 1990; Jacobson et al. 1956; Snellings et al. 1984a), and normochromic anemia occurred in dogs at 102 ppm for six months (Jacobson et al. 1956). Splenic extramedullary hematopoiesis and focal fibrosis were observed in rats at 50–100 ppm for 104 weeks, again in the setting of concurrent colony infection (Lynch et al. 1984a, 1984b). No hematological effects occurred in mice at up to 100 ppm for 102 weeks (NTP 1987) or in gavaged rats at up to 30 mg/kg/day (Hollingsworth et al. 1956).

Musculoskeletal. Muscular atrophy was reported in rats, rabbits, monkeys, and dogs exposed at 204–357 4 ppm for six weeks to 226 days (Hollingsworth et al. 1956; Jacobson et al. 1956), and multifocal skeletal myopathy occurred in rats at 100 ppm for 104 weeks (NOAEL 50 ppm) (Lynch et al. 1984a, 1984b).

Hepatic. Slight discoloration and fatty degeneration of the liver occurred in rats and guinea pigs after brief exposure to 841 ppm (Hollingsworth et al. 1956), and an elevated liver-to-body-weight ratio (without his tological change) occurred in female mice at 250 ppm for 11 weeks (Snellings et al. 1984a). No hepatic effects were seen in rats or mice at 100 ppm for up to 104 weeks (Lynch et al. 1984a, 1984b; NTP 1987; EPA 1994). By gavage, slight liver damage occurred at 100 mg/kg/day but not at 30 mg/kg/day (Hollingsworth et al. 1956).

Renal. Tubular enlargement, congestion, and cloudy swelling occurred in rats and guinea pigs after brief exposure to 841 ppm (Hollingsworth et al. 1956); renal tubular degeneration and necrosis occurred in mice at 100–200 ppm for 14 weeks, with more severe necrosis at lethal concentrations (400–600 ppm) (NTP 1987). No renal effects occurred at 50–100 ppm for up to 104 weeks (Lynch et al. 1984a, 1984b; NTP 1987).

Dermal and Ocular. Concentrated EtO produces a freezing/frostbite-like injury on skin contact, ranging from first- to third-degree burns (Taylor 1977), and dermal application of 10–50% aqueous solutions to rabbits produced hyperemia and edema, with undiluted EtO causing severe erythema, subdermal hemorrhage, and chemical burns (Celanese Chem Co. 1972; Hollingsworth et al. 1956). Human case reports describe erythema, blistering, and second-degree burns from EtO-sterilized medical supplies (Alomar et al. 1981). Ocular instillation of a 1% solution produced slight corneal cloudiness in rabbits (McDonald et al. 1977), and cataracts occurred in monkeys (2/11–3/12 vs. 0/11 controls) and rats (3/79–9/78 vs. 2/77 controls) after two years of intermittent exposure at 50–100 ppm (Lynch et al. 1984a, 1984b)—a finding that parallels human case reports of cataracts in sterilizer operators acutely exposed to an estimated ≥700 ppm (Gross et al. 1979; Jay et al. 1982).

Endocrine. Pale, enlarged adrenals with cortical fat vacuolation occurred in rats and guinea pigs after brief exposure to 841 ppm (Hollingsworth et al. 1956), and adrenal cortical vacuolation and hyperplasia occurred in rats at 50 ppm for up to 104 weeks, again confounded by colony infection (Lynch et al. 1984a, 1984b). No thyroid, parathyroid, adrenal, or pituitary effects occurred in mice at up to 100 ppm for 102 weeks (NTP 1987). ATSDR’s systematic review classified the endocrine system as a suspected human health effect target based on this animal database, since no human endocrine data were located.

Immunological. Thymic lymphocytic hypoplasia occurred in mice at 400 ppm for up to two weeks, with necrosis of thymus and spleen at 600 ppm (NTP 1987). In humans, dermal EtO exposure has been repeat edly implicated as a contact sensitizer, producing allergic and delayed-type hypersensitivity dermatitis in case reports of exposed healthcare workers and patients (Belen and Polat 2015; Brashear et al. 1996; Car oli et al. 2005; Dagregorio and Guillet 2004; Kerre and Goossens 2009; Lerman et al. 1995; Romaguera and Vilaplana 1998), consistent with earlier experimental sensitization findings in volunteers (Sexton and Henson 1950; Shupack et al. 1981).

Neurological. Neurotoxicity is among the most consistently documented noncancer effects across species. In animals, a single exposure produced incoordination and semiconsciousness in mice at 1,600 ppm and decreased motor activity in rats at 300–500 ppm (NTP 1987; EPA 2005a). Repeated exposure at 100–500 ppm for 48–226 days produced impaired hindlimb sensorimotor function, decreased grip strength, altered gait, and peripheral neuropathy in rats, mice, and other species (Hollingsworth et al. 1956; Jacobson et al. 1956; Kaido et al. 1992; Matsuoka et al. 1990; Ohnishi et al. 1985). In a nine-month study, rats exposed at 250 ppm showed retarded myelination and mild axonal degeneration in hindleg nerves even without overt clinical signs (Ohnishi et al. 1986), and mechanistic follow-up demonstrated a 33% reduction in ax onal transport velocity together with increased myelinated-fiber degeneration in rats exposed at 500 ppm (Nagata et al. 1992). Demyelination and axonal dystrophy were also reported in the brains of chronically exposed monkeys, albeit in a small-sample study with a background lesion in one control animal (Lynch et al. 1984a). No neurological effects occurred in a two-generation rat study at up to 100 ppm (EPA 1994) or in chronically exposed mice at up to 100 ppm (NTP 1987), suggesting a dose threshold below approximately 100 ppm for overt peripheral neuropathy in these study designs. In humans, case series of chronically exposed workers describe peripheral neuropathy, impaired hand-eye coordination, cognitive dysfunction, memory loss, and hand numbness at estimated average exposures as low as 3 ppm (with higher short-term peaks), with sural nerve biopsies confirming axonal degeneration and regeneration (Brashear et al. 1996; Crystal et al. 1988; Dretchen et al. 1992; Estrin et al. 1987; Finelli et al. 1983; Kuzuhara et al. 1983; Sali nas et al. 1981; Schröder et al. 1985; Zampollo et al. 1984), though these case reports lack the controlled dose-response design of the animal studies.

Reproductive. In female rats exposed at 150 ppm for three weeks premating and through gestation day 16, increased resorptions and decreased pups per litter occurred (NIOSH 1982), and a two-generation study found increased post-implantation loss at 33 ppm and decreased live pups per litter at 100 ppm across F1 and F2 generations (EPA 1994). Male reproductive toxicity was demonstrated across several studies: de creased testicular and epididymal weight, reduced germ cell survival and sperm count, and seminiferous tubule histopathology occurred in rats exposed at 250–500 ppm for 6–13 weeks (Kaido et al. 1992; Mori et al. 1991a), an effect partially mitigated by methylcobalamin co-treatment in a mechanistic follow-up (Mori et al. 1991b). In monkeys, 24 months of exposure at 50 ppm (the lowest level tested) reduced sperm count by 28% and motility by 32%, though functional fertility was not directly assessed (Lynch et al. 1984a). In hu mans, epidemiological studies of sterilizer workers have explored associations with spontaneous abortion; Hemminki et al. (1982) reported abortion rates of 15.1% in EtO-exposed versus 4.6% in unexposed Finnish hospital workers, Rowland et al. (1996) found an adjusted relative risk of 2.5 (95% CI 1.0–6.3) for spon taneous abortion among dental sterilizer operators, and Gresie-Brusin et al. (2007) reported a prevalence odds ratio of 20.8 (95% CI 2.1–199.3) for spontaneous abortion among South African sterilizer operators; all three studies, however, are limited by small sample sizes, absence of quantitative exposure measurement, and reliance on self-reported data.

Developmental. Reduced fetal body weight, shortened crown-rump length, and increased skeletal ossi fication deficits occurred in rat offspring following maternal exposure at 100–1,200 ppm across several studies (Neeper-Bradley and Kubena 1993; NIOSH 1982; Saillenfait et al. 1996; Snellings et al. 1982a), and decreased pup body weight occurred at 33–100 ppm in a two-generation design (EPA 1994). Saillenfait et al. (1996) additionally reported increased renal pelvis and ureteral dilation in fetuses at 1,200 ppm. In mice, a single very brief (1.5-hour) high-concentration exposure (1,200 ppm) at defined times after mating produced fetal defects, predominantly hydrops and ocular malformations (Rutledge and Generoso 1989), implicating a narrow window of heightened conceptus sensitivity. No developmental effects occurred in rabbits exposed at 150 ppm during gestation (NIOSH 1982). No human developmental studies were located.

5. Animal Carcinogenicity Bioassay Findings

EtO has produced tumors at multiple sites across at least three chronic rodent bioassays and several shorter term strain-specific assays, generally at exposure levels that also produced noncancer toxicity.

In the pivotal two-year inhalation bioassay in Fischer 344 rats (NIOSH-sponsored; reported across several companion publications), male rats exposed at 100 ppm showed significantly increased incidence of brain gliomas (5/79 vs. 0/76 controls) and peritoneal mesotheliomas (21/79 vs. 3/78 controls), and mononuclear cell leukemia was significantly elevated at 50 ppm (38/79 vs. 24/77 controls) (Lynch et al. 1984a, 1984b). A related analysis of the same rat cohort reported spleen mononuclear cell leukemia significantly elevated in females at 100 ppm (58% vs. 8–11% in controls), with non-significant dose-related increases also apparent at 10 and 33 ppm, plus significant mortality-adjusted trends for mononuclear cell leukemia at exposures of 33 ppm and above, subcutis fibroma, and primary brain neoplasms in both sexes (Garman et al. 1985, 1986; Snellings et al. 1984b).

In a two-year inhalation bioassay in B6C3F1 mice, males exposed at 100 ppm developed significantly in creased alveolar/bronchiolar carcinoma (16/50 vs. 6/50 controls), and females developed significantly in creased alveolar/bronchiolar adenoma (17/49 vs. 2/49) and carcinoma (7/49 vs. 0/49) (NTP 1987). Harde rian gland papillary cystadenoma incidence was elevated at both 50 and 100 ppm in both sexes, and female mice at 100 ppm also showed marginally increased malignant lymphoma (22/49 vs. 9/49) and uterine ade nocarcinoma (5/49 vs. 0/49); female mice at 50 ppm (but not 100 ppm) showed increased hepatocellular adenoma and mammary gland adenocarcinoma/adenosquamous carcinoma (NTP 1987). An independent pathology re-review confirmed these tumor findings (Picut et al. 2003).

A strain-specific inhalation bioassay in female A/J mice, a strain especially susceptible to lung tumorigenesis, found significantly increased lung adenoma incidence after six months of exposure at 70 or 200 ppm (16/28 and 25/29, respectively, vs. 8/30 in controls), with results replicated in an independent cohort (Adkins et al. 1986). By the oral route, dose-related increases in malignant forestomach tumors (predominantly squa mous cell carcinoma) occurred in female rats gavaged at 7.5 or 30 mg/kg/day twice weekly for 150 weeks (8/50 and 31/50 tumor-bearing animals, respectively, vs. none in 100 combined controls), consistent with a local, high-concentration contact effect at the site of administration (Dunkelberg 1982). A lifetime der mal skin-painting study in mice using a 10% EtO solution did not produce skin tumors, suggesting the car cinogenic hazard is most pronounced via routes producing sustained internal or respiratory-tract exposure rather than superficial dermal contact (Van Duuren et al. 1965).

Collectively, these bioassays span three genetically distinct rodent models (Fischer 344 rats, B6C3F1 mice, and A/J mice) and implicate lymphohematopoietic tissue, brain, peritoneum, lung, and, in mice, mammary gland and uterus as target sites—an unusually broad multi-site tumor profile consistent with a genotoxic, DNA-reactive mode of carcinogenic action rather than a site-restricted, receptor-mediated, or purely irritant mechanism.

6. Comparison of Animal Findings to Human Epidemiological Evidence

Human evidence for EtO carcinogenicity comes almost entirely from occupational cohorts of production and sterilization workers. The National Institute for Occupational Safety and Health (NIOSH) cohort—the largest and most extensively analyzed—found positive exposure-response trends for both lymphohematopoietic cancers and female breast cancer with increasing cumulative exposure (Steenland et al. 2003, 2004). Meta analyses of the broader occupational literature have reached somewhat divergent conclusions depending on methodology and included studies: an early meta-analysis found no significant excess of leukemia or non Hodgkin’s lymphoma (Shore et al. 1993), largely unchanged in a later update (Teta et al. 1999), whereas a more recent systematic review and meta-analysis reported a significant overall excess of lymphohematopoietic cancer (meta-relative risk 1.48, 95% CI 1.07–2.05) but no excess of breast cancer (meta-relative risk 0.97, 95% CI 0.80–1.18), with higher risks concentrated in earlier, methodologically weaker studies (Marsh et al. 2019). Individual cohort and case-cluster studies have reported excess leukemia and lymphohematopoi etic cancer in specific sterilization facilities (Hogstedt et al. 1979a, 1986) and excess breast cancer in a small nursing cohort (Tompa et al. 1999), alongside numerous cohort mortality studies with mixed or null find ings for various cancer sites (Bisanti et al. 1993; Coggon et al. 2004; Gardner et al. 1989; Hagmar et al. 1991, 1995; Kiesselbach et al. 1990; Mikoczy et al. 2011; Morgan et al. 1981; Norman et al. 1995; Olsen et al. 1997; Swaen et al. 2009; Teta et al. 1993; Thiess et al. 1981; Wong and Trent 1993).

The concordance between animal and human evidence is strongest, and most mechanistically coherent, for lymphohematopoietic malignancies: both the rat bioassay (mononuclear cell leukemia) and multiple human cohorts point to the hematopoietic/lymphoid system as a sensitive target, a convergence regula tory bodies have cited as key support for classifying EtO as a human carcinogen (the U.S. National Toxicology Program lists EtO as “known to be a human carcinogen,” EPA has characterized it as “carcinogenic to humans” by inhalation, and IARC has designated it Group 1, carcinogenic to humans). The evidence for breast cancer is less consistent between the two databases: mouse bioassays show mammary tumors at the lower (50 ppm) but not higher (100 ppm) dose in one study (NTP 1987), a non-monotonic pattern that complicates simple exposure-response inference, while human data are similarly mixed across cohorts and meta-analyses (Marsh et al. 2019; Steenland et al. 2003). For solid tumor sites prominent in rodent bioassays—brain glioma, peritoneal mesothelioma, and lung tumors—human epidemiological support is comparatively weak or absent, despite decades of cohort follow-up.

Several toxicokinetic and study-design factors likely contribute to this animal-human divergence. First, species differences in metabolic partitioning are substantial: PBPK modeling indicates that rats and mice route a much larger share of absorbed EtO through glutathione conjugation (a detoxification pathway) than do humans, who rely more heavily on hydrolysis (Fennell and Brown 2001); this does not straightforwardly predict lower internal dose in humans, since simulated peak blood concentrations and AUC were, if any thing, somewhat higher in humans than in rodents at equivalent external exposure, but it does mean that internal dose-metrics derived from rodent bioassays require model-based, rather than simple ppm-for-ppm, extrapolation to humans. Second, rodent bioassays used sustained, high, and relatively uniform exposures (10–100+ ppm for up to two years), whereas human cohorts experienced variable, often intermittent, and imprecisely reconstructed historical exposures, frequently with substantial exposure misclassification—a limitation explicitly acknowledged for the spontaneous abortion studies (Gresie-Brusin et al. 2007; Hemminki et al. 1982; Rowland et al. 1996) and for many cancer cohorts, where exposure levels were rarely measured directly. Third, statistical power in human cohorts is constrained by the rarity of most cancer endpoints and by the “healthy worker effect,” a well-recognized bias in which employed cohorts show lower baseline mortality than the general population, tending to mask true excess risk. Fourth, some rodent tu mor sites (mouse mammary gland, mouse uterus, rat brain and peritoneum) may reflect species-specific susceptibilities, such as background tumor rates and hormonal milieu differences, that do not directly gen eralize even where the underlying genotoxic mechanism is shared. Finally, some early human cohorts (particularly chlorohydrin-process workers) had co-exposures to other genotoxic chemicals, complicating attribution of observed cancer excesses specifically to EtO, whereas rodent bioassays used purified EtO exposures without such confounding.

Overall, regulatory agencies have concluded that the combination of consistent genotoxicity across species (including biomarker evidence of chromosomal damage in exposed workers), multi-site rodent carcinogenic ity, and epidemiological evidence—strongest for lymphohematopoietic cancer—collectively supports a mu tagenic mode of carcinogenic action operating similarly across species, even where individual tumor sites or magnitude of risk do not map one-to-one between animal bioassays and human cohort findings.

7. Data Gaps and Areas of Ongoing Research

Several evidence gaps constrain confidence in specific aspects of the EtO toxicological database. Chronic duration noncancer toxicity data are notably inadequate: the principal chronic rat inhalation bioassay was confounded by a concurrent bacterial infection affecting all treatment groups including controls, under mining interpretation of chronic adrenal, hematological, and respiratory findings, and no adequate human exposure-response data exist for a chronic inhalation reference level. Well-controlled, long-term inhalation studies in disease-free animal colonies, evaluating a comprehensive battery of noncancer endpoints below carcinogenic thresholds, would meaningfully strengthen chronic risk assessment.

Mechanistic gaps remain for several noncancer endpoints. The proximate toxicant(s) responsible for EtO-induced neurotoxicity have not been fully characterized, and further mechanistic work—building on the axonal transport findings of Nagata et al. (1992)—could support physiologically based dose-response modeling for neurological endpoints specifically, rather than relying on default uncertainty factors. Adrenal/endocrine effects have been observed in only a small number of studies, one confounded by infection, leaving the endocrine hazard identification (currently categorized as “suspected” rather than “presumed” for humans) comparatively uncertain. Reproductive toxicity data, while suggestive of male reproductive organ effects and adverse pregnancy outcomes, would benefit from a dedicated, comprehen sive inhalation reproductive toxicity study designed to establish clearer exposure-response relationships, since existing animal reproductive studies were not originally designed for that purpose.

Toxicokinetic data gaps include limited characterization of EtO’s blood half-life in relation to biological moni toring, and near-total absence of oral or dermal absorption, distribution, and excretion data—a gap of mod est practical significance given that inhalation dominates real-world exposure, but one that limits assessment of accidental oral/dermal scenarios. Comparative PBPK modeling has advanced considerably (Fennell and Brown 2001; Krishnan et al. 1992), but identification of the specific proximate toxicant(s) driving non cancer effects remains necessary before these models can be fully applied to interspecies extrapolation for endpoints beyond cancer.

Children’s susceptibility remains largely unstudied directly: no age-stratified toxicity or carcinogenicity data exist, and inferences about heightened pediatric susceptibility rest on indirect evidence of immature detox ification enzyme ontogeny (McCarver and Hines 2002; Zhong et al. 2018) rather than direct developmental toxicology. Environmental fate data are also dated: current measurements of atmospheric and aquatic degradation rates are sparse, many analyses still rely on decades-old rate constants, and no soil degrada tion studies have been located, limiting confidence in modeled persistence estimates. Human biomonitor ing continues to evolve, and the NIOSH occupational cohort continues to be followed prospectively, which should improve statistical power over time for cancer endpoints with long latency and help resolve incon sistencies between earlier and more recent cohort analyses.

Bibliography