Benzene Risk Assessment and Leukemia and Dose Response

RoseyFields conducts benzene risk assessments using the following methodology, which integrates chemical-specific toxicokinetic and toxicological data, established regulatory dose-response values, and site- or case-specific exposure information to characterize potential human health risks associated with benzene exposure. This methodology is applied consistently across regulatory, community exposure, and toxic-tort/FELA litigation matters, and it is designed to withstand scrutiny under Daubert and analogous evidentiary standards because it relies on peer-reviewed, transparently derived dose-response values and a documented, stepwise application of standard EPA risk assessment equations.

1. Chemical Identity, Regulatory Background, and Exposure Sources

Benzene (CAS No. 71-43-2), also known as benzol, is a volatile aromatic hydrocarbon with the empirical formula C6H6 and a molecular weight of 78.11 g/mol. It is a clear, colorless to light-yellow liquid at room temperature with a characteristic aromatic odor, a vapor pressure of approximately 75 mm Hg at 20°C, and a boiling point of 80.1°C. Its water solubility of roughly 1,750 mg/L, Henry’s law constant of approximately 5.5 × 10⁻³ atm-m³/mol, and log Koc of approximately 1.8–1.9 make benzene both appreciably water-soluble and highly volatile, properties that govern its fate and transport in soil, groundwater, and air and that make it a frequent driver of vapor-intrusion and air-dispersion evaluations.

Benzene has been recognized as a hematotoxicant for more than a century and remains one of the most heavily regulated industrial chemicals in the United States. OSHA has established a permissible exposure limit of 1 ppm as an 8-hour time-weighted average, with a short-term excursion limit of 5 ppm, reflecting the historical occupational hematotoxicity data underlying this methodology. The ACGIH threshold limit value is 10 ppm (8-hour TWA). On the environmental side, EPA’s Integrated Risk Information System (IRIS) provides both noncancer toxicity values — an inhalation reference concentration (RfC) and an oral reference dose (RfD) — and a separate quantitative cancer assessment addressing benzene’s classification as a known human carcinogen. RoseyFields’ methodology, described below, follows the dose-response derivation logic and underlying peer-reviewed studies that support these regulatory values, while independently verifying the toxicological literature that underlies each value before applying it in a specific engagement.

Benzene is ubiquitous in the industrial and ambient environment. Historically, and in the present day, principal sources and pathways of human exposure include:

• Petroleum refining and petrochemical manufacturing, where benzene is both a feedstock and a byproduct, and where fugitive emissions, tank venting, and flaring are common airborne release mechanisms evaluated using air dispersion models such as AERMOD;
• Gasoline and other refined fuels, in which benzene is present as a natural constituent (historically at higher concentrations before regulatory limits reduced permissible benzene content), giving rise to exposures at fueling stations, along transportation corridors, and near bulk storage and pipeline facilities;
• Combustion sources, including vehicle exhaust and tobacco smoke, both of which are well-documented contributors to ambient and indoor benzene concentrations;
• Industrial solvent use, historically including benzene’s use as a general laboratory and industrial solvent, a rubber-industry component (notably in the historical Pliofilm and shoe manufacturing industries), and a chemical intermediate;
• Contaminated soil and groundwater at refinery, bulk terminal, pipeline, and manufactured gas plant sites, which can result in vapor intrusion into overlying structures or incidental ingestion/dermal contact pathways where groundwater serves as a drinking water source; and
• Occupational settings involving direct handling of benzene-containing products, including chemical manufacturing, petroleum refining, rubber and shoe manufacturing, and rail-related fuel and chemical-transport operations relevant to FELA claims.

The dominant route of environmental and occupational exposure is inhalation, owing to benzene’s high volatility, though incidental ingestion of contaminated groundwater and, to a much lesser extent, dermal contact are also evaluated where relevant to a specific exposure scenario. Dermal absorption of benzene is comparatively minor because rapid evaporation from the skin surface limits the quantity available for percutaneous uptake, and gastrointestinal absorption of both aqueous and oil-based benzene has been shown in animal mass-balance studies to exceed 90% of the administered dose (Parke and Williams, 1953; Sabourin et al., 1987).

2. Point-of-Departure (POD) Selection

RoseyFields’ benzene methodology follows the point-of-departure framework used in EPA’s peer-reviewed chronic noncancer toxicity assessment, which is itself grounded in the underlying primary literature rather than in agency judgment alone. The principal study selected for derivation of both the chronic inhalation RfC and the chronic oral RfD is the human occupational cross-sectional study of Rothman et al. (1996a), which compared hematologic parameters in 44 workers occupationally exposed to benzene with those of 44 age- and gender-matched unexposed controls in Shanghai, China. Benzene exposure was individually monitored using organic vapor passive dosimetry and corroborated with urinary benzene metabolite measurements, and co-exposure to other solvents (notably toluene) was minimal, substantially reducing the potential for confounding relative to many other occupational cohorts.

Rothman et al. (1996a) reported statistically significant reductions in absolute lymphocyte count (ALC), white blood cell (WBC) count, red blood cell (RBC) count, and platelet count, and a significant increase in mean corpuscular volume (MCV), in the high-exposure group (median 8-hour time-weighted average of 91.9 ppm). At lower exposures (median 13.6 ppm), ALC, RBC, and platelet counts remained significantly reduced relative to controls, and a subgroup exposed to a median of 7.6 ppm still exhibited a statistically significant reduction in ALC (p < 0.03). This dose-response gradient, extending to comparatively low exposure concentrations, makes the study well-suited to benchmark dose (BMD) modeling and identifies ALC depression as the most sensitive quantifiable endpoint. ALC reduction is treated as a sentinel biomarker for benzene hematotoxicity because it correlates inversely with urinary benzene metabolite levels and because analogous lymphocyte depression has been linked, in more heavily exposed cohorts, to the pattern of bone marrow suppression that precedes clinically apparent benzene poisoning (Dosemeci et al., 1996).

Two supporting/comparison studies are used to assess the consistency of the human-data-derived values against experimental animal data. For the inhalation pathway, the subchronic study of Ward et al. (1985) exposed CD-1 mice and Sprague-Dawley rats to 0, 1, 10, 30, or 300 ppm benzene for 6 hours/day, 5 days/week, for 91 days. The male mouse was the most sensitive sex/species, with a NOAEL of 30 ppm and a LOAEL of 300 ppm for hematologic effects; decreased hematocrit (HCT) was selected as the critical effect for BMD modeling. For the oral pathway, the chronic gavage bioassay conducted by the National Toxicology Program (NTP, 1986) in F344 rats and B6C3F1 mice (103 weeks, both sexes, doses ranging from 25 to 200 mg/kg depending on sex and species) identified a LOAEL of 25 mg/kg for lymphocytopenia and leukopenia in female rats and in both sexes of mice, and 50 mg/kg in male rats; no NOAEL was identified because these were the lowest doses tested. Additional shorter-duration studies — including the subchronic drinking-water studies of Hsieh et al. (1988b) and White et al. (1984), and the low-concentration inhalation studies of Baarson et al. (1984) and Cronkite et al. (1985) — provide independent confirmation that statistically significant hematologic effects occur at exposure levels considerably below the NTP and Ward et al. LOAELs, supporting the selection of a sensitive point of departure rather than relying on a single study’s NOAEL/LOAEL.

3. Pharmacokinetic Modeling and Dose Conversion

Because the principal study (Rothman et al., 1996a) reports occupational, intermittent (8-hour/day, 5-day/week) inhalation exposures, RoseyFields’ methodology performs a series of standard dosimetric adjustments before the point of departure can be used to derive continuous, chronic-duration toxicity values applicable to environmental or 24-hour occupational-equivalent exposure scenarios.

First, benchmark concentrations (in ppm) are converted to mg/m³ using benzene’s molecular weight (78.11 g/mol) and the standard molar volume of 24.45 L/mol at 25°C and 760 mm Hg. Second, the resulting value is adjusted from an occupational exposure pattern to a continuous ambient exposure pattern using the ratio of default occupational to ambient minute ventilation volumes (10 m³ over an 8-hour workday versus 20 m³ over 24 hours) and the ratio of a 5-day to a 7-day exposure week. This dosimetric adjustment methodology follows EPA’s standard inhalation dosimetry approach (U.S. EPA, 1994) and, for the interspecies comparison using Ward et al. (1985), assumes ppm equivalence across species (a regional gas dose ratio of 1.0), appropriate for a systemically acting vapor such as benzene that is not primarily a portal-of-entry (respiratory tract) toxicant.

Route-to-route (inhalation-to-oral) extrapolation is required because no adequate human hematotoxicity data are available for the oral pathway. This extrapolation relies on the relative absorption efficiency of benzene by the two routes: an inhalation absorption fraction of approximately 50% and an oral absorption fraction of approximately 100%. The inhalation absorption value is supported by a consistent body of human respiratory retention/uptake studies, including Srbova et al. (1950), who reported absorption ranging from roughly 20–50% after the first hour of exposure in human subjects at 47–100 ppm; Nomiyama and Nomiyama (1974), who reported average respiratory uptake of approximately 47% in six human subjects exposed at 52–62 ppm; and Pekari et al. (1992), who reported average absorption of 48–52% in three male subjects exposed at 1.7 and 10 ppm. The oral absorption value is supported by the mass-balance gavage study of Sabourin et al. (1987) in rats and mice, which demonstrated gastrointestinal absorption exceeding 97% across a dose range of 0.5–150 mg/kg, and is broadly consistent with the earlier rabbit mass-balance study of Parke and Williams (1953), which recovered approximately 90% of an orally administered radiolabeled dose as metabolites, CO2, and exhaled unchanged benzene.

Using these absorption fractions, an equivalent oral dose rate is calculated by multiplying the dosimetrically adjusted inhalation concentration by a default inhalation rate (20 m³/day), applying a factor of 0.5 to correct for the higher relative efficiency of oral versus inhalation absorption, and dividing by the default adult body weight of 70 kg. This same absorption-ratio approach is used by EPA for route-to-route extrapolation of benzene’s cancer risk estimates and is applied identically here for the noncancer hematotoxicity endpoint.

Interindividual physiological variability in benzene toxicokinetics is also considered qualitatively at this stage, even though it is not incorporated as a separate numerical adjustment beyond the standard intraspecies uncertainty factor discussed below. Physiologically based pharmacokinetic (PBPK) modeling by Brown et al. (1998), using empirical data from Sato et al. (1975), demonstrated that women metabolize approximately 23–26% more benzene than men under identical exposure conditions, despite generally lower peak blood benzene concentrations in women, reflecting differences in body fat content and metabolic capacity. These physiological differences are noted as a source of variability that is addressed, at a population level, through the intraspecies (human variability) uncertainty factor rather than through route-specific or sex-specific point estimates.

4. Uncertainty Factor Derivation

Consistent with EPA’s standard noncancer risk assessment framework, RoseyFields applies a composite uncertainty factor (UF) to the dosimetrically adjusted point of departure to derive each toxicity value. The composite UF for the primary (Rothman et al., 1996a-based, BMD-derived) RfC and RfD is 300, composed of the following individual factors:

• Effect-level extrapolation factor (analogous to a LOAEL-to-NOAEL UF) of 3: applied because the benchmark concentration/dose (BMC/BMD), while statistically derived from a full dose-response curve, still represents an adverse-effect level rather than a no-effect level. A factor of 3, rather than the full default of 10, is used because ALC depression, while a genuine adverse effect and biologically meaningful sentinel marker, is not itself a frank, functionally impairing effect at the benchmark response level selected (one standard deviation change from the control mean, corresponding to an approximate 10% excess risk in the tails of a normal distribution).
• Interspecies extrapolation factor: not applicable for the human-data-based values (the Rothman et al. study is a human epidemiologic study), but a factor of 3 (inhalation) or 10 (oral) is applied for the corresponding animal-based comparison values (Ward et al., 1985; NTP, 1986), reflecting toxicokinetic and toxicodynamic uncertainty in extrapolating from rodents to humans.
• Intraspecies (human variability) factor of 10: applied uniformly to protect sensitive human subpopulations, reflecting documented genetic polymorphisms in benzene-metabolizing enzymes (CYP2E1, NAD(P)H:quinone oxidoreductase [NQO1], glutathione S-transferases) that materially affect individual susceptibility. Rothman et al. (1997) demonstrated that Chinese benzene-exposed workers with both rapid CYP2E1 activity and a homozygous variant NQO1 genotype had an approximately 7.6-fold increased risk of hematotoxicity compared with workers with slow CYP2E1 activity and wild-type NQO1 alleles, and Rossi et al. (1999) similarly found genetic polymorphisms at multiple loci contributing to variability in benzene metabolite excretion among occupationally exposed bus drivers.
• Subchronic-to-chronic extrapolation factor: a partial factor of 3 (rather than the full default of 10) is applied to the human-data-based values because the mean occupational exposure duration in the Rothman et al. (1996a) cohort (6.3 years, range 0.7–16 years) is close to, though somewhat short of, the 7-year benchmark conventionally used to distinguish subchronic from chronic human exposure. This factor is not applicable to the NTP (1986) comparison value because that study’s 103-week gavage duration is already chronic.
• Database uncertainty factor of 3: applied uniformly to reflect the absence of a two-generation reproductive/developmental toxicity study in the overall benzene toxicological database, notwithstanding the otherwise extensive body of hematotoxicity, reproductive, developmental, and mechanistic literature.

The composite UF for the animal-data-based comparison values is 1000 (inhalation, Ward et al., 1985: 3 × 3 × 10 × 3 × 3; oral, NTP, 1986: 3 × 10 × 10 × 3), reflecting the additional interspecies extrapolation factor not required for the human-data-based values. A summary of the applied uncertainty factors is provided in Table 1.

5. Derivation of the Oral Reference Dose (RfD) and Inhalation Reference Concentration (RfC)

Applying the point-of-departure and uncertainty factor framework described above, RoseyFields’ methodology derives the following chronic toxicity values for benzene, consistent with the peer-reviewed dose-response literature discussed above:

Chronic Inhalation Reference Concentration (RfC): The dosimetrically adjusted benchmark concentration lower confidence limit (BMCLADJ) of 8.2 mg/m³, derived from BMD modeling of the Rothman et al. (1996a) ALC data, divided by the composite UF of 300, yields a chronic inhalation RfC of 3 × 10⁻² mg/m³. This value is supported by comparison values derived from the same study’s LOAEL (9 × 10⁻³ mg/m³, UF of 1000) and from the Ward et al. (1985) rodent study using both BMD modeling (5 × 10⁻² mg/m³, UF of 1000) and the NOAEL/UF approach (6 × 10⁻² mg/m³, UF of 300). All four estimates fall within a single order of magnitude (9 × 10⁻³ to 6 × 10⁻² mg/m³), providing reasonable confidence in the selected value.

Chronic Oral Reference Dose (RfD): The equivalent oral dose rate of 1.2 mg/kg-day, obtained via route-to-route extrapolation of the same Rothman et al. (1996a) BMCLADJ, divided by the composite UF of 300, yields a chronic oral RfD of 4 × 10⁻³ mg/kg-day. This value is supported by a LOAEL-based comparison value from the same study (1 × 10⁻³ mg/kg-day, UF of 1000) and by values derived from the NTP (1986) rodent bioassay using BMD modeling (7 × 10⁻⁴ mg/kg-day, UF of 1000) and the LOAEL/UF approach (6 × 10⁻³ mg/kg-day, UF of 3000). These values likewise converge within roughly one order of magnitude (7 × 10⁻⁴ to 6 × 10⁻³ mg/kg-day), reinforcing confidence in the selected RfD.

A summary of the RfC and RfD values derived by each approach and dataset is presented in Table 2. RoseyFields’ assessors select the human-data-based, BMD-derived values (RfC = 3 × 10⁻² mg/m³; RfD = 4 × 10⁻³ mg/kg-day) as the primary toxicity values for hazard quotient and hazard index calculations, because they are derived from good-quality human epidemiologic data that avoid the interspecies extrapolation uncertainty inherent in rodent-based values, and because BMD modeling makes fuller use of the available dose-response data than a single-dose NOAEL/LOAEL approach.

6. Carcinogenicity Classification and Cancer Risk Values

The dose-response methodology described in Sections 2 through 5 addresses benzene’s noncancer (principally hematotoxic) endpoints. Benzene’s carcinogenicity assessment is derived separately, reflecting the distinct epidemiologic and mechanistic evidence base underlying its cancer classification. Benzene has long been recognized as a human carcinogen on the basis of a consistent and extensive epidemiologic literature linking occupational exposure to an excess incidence of acute myelogenous leukemia and related hematologic malignancies. The NIOSH Pliofilm rubber-worker cohort, analyzed by Rinsky et al. (1987), remains one of the principal epidemiologic datasets supporting this classification, having demonstrated a significant excess of leukemia mortality among rubber-hydrochloride workers with quantified historical benzene exposure. This and related occupational cohort literature — together with the extensive genotoxicity and mechanistic evidence discussed in Section 8, including the demonstrated role of CYP2E1-mediated metabolic activation (see Section 3) and benzene metabolite-induced chromosomal damage — underlie EPA’s quantitative cancer risk assessment for benzene, which is documented separately from the noncancer toxicity values addressed in this methodology.

Because the derivation of a specific cancer slope factor (CSF) or inhalation unit risk (IUR) estimate involves separate epidemiologic dose-reconstruction and modeling choices, RoseyFields verifies and applies the specific, current CSF/IUR value from the applicable regulatory source (e.g., the current EPA IRIS chemical assessment summary, or an applicable state agency value where required by the jurisdiction) at the time of each engagement, rather than relying on a fixed historical figure. When cancer risk is a component of an assessment, RoseyFields applies the verified IUR to reconstructed or modeled air concentrations, and the verified oral CSF (derived via the same route-to-route extrapolation logic described in Section 3) to reconstructed oral intake estimates, to calculate excess lifetime cancer risk as described in Section 7 below.

7. Application in Practice: Hazard Quotients, Cancer Risk, and Exposure Reconstruction

RoseyFields’ risk assessors apply the RfC, RfD, and, where applicable, cancer risk values described above through the standard EPA risk characterization framework:

Hazard Quotient (HQ) and Hazard Index (HI): For a given exposure pathway, the hazard quotient is calculated as the estimated chronic daily intake (for oral exposure) or air concentration (for inhalation exposure) divided by the corresponding RfD or RfC. An HQ at or below 1.0 indicates that the estimated exposure is at or below a level associated with an appreciable risk of noncancer adverse effects, incorporating the built-in uncertainty and safety margins reflected in the composite UF. Where benzene exposure occurs concurrently with other chemicals affecting the same target organ system (in this case, the hematopoietic/bone marrow system), individual chemical HQs are summed to a hazard index (HI) to evaluate the potential for cumulative or additive hematotoxicity.

Excess Lifetime Cancer Risk (ELCR): Chronic daily intake or air concentration estimates are multiplied by the applicable oral CSF or inhalation IUR to calculate an incremental probability of developing cancer over a lifetime attributable to the exposure being evaluated, consistent with standard EPA risk assessment guidance.

Exposure and Dose Reconstruction: In litigation and regulatory matters, RoseyFields typically must first reconstruct historical or current airborne benzene concentrations, or groundwater/soil concentrations relevant to an ingestion or vapor-intrusion pathway, before HQ/HI or ELCR calculations can be performed. This reconstruction commonly relies on AERMOD air dispersion modeling calibrated to facility-specific emissions data, historical industrial hygiene or ambient air monitoring records, published emission factors for the source category at issue (e.g., refinery fugitive emissions, rail yard fueling operations, or bulk terminal operations), or a combination of these approaches. Chronic daily intake is then calculated using the standard EPA exposure equation — concentration multiplied by intake or contact rate, exposure frequency, and exposure duration, divided by body weight and averaging time — parameterized with case-specific or default exposure factors as appropriate to the claimant’s documented activity pattern.

Litigation and Regulatory Support: This methodology supports causation, dose-response, and general-and-specific-causation opinions in toxic tort and FELA litigation involving alleged benzene-induced hematologic disease (including aplastic anemia, myelodysplastic syndrome, and acute myelogenous leukemia), refinery and petrochemical facility fenceline exposure claims, and community exposure matters involving contaminated groundwater or vapor intrusion. Because the underlying dose-response values are derived transparently from peer-reviewed human and animal studies using a documented, standard methodology (BMD modeling, EPA inhalation dosimetry, and route-to-route extrapolation, each traceable to specific published studies), the resulting opinions are structured to satisfy the reliability and “fit” requirements of Daubert and analogous evidentiary standards, with methodologies evaluated under Daubert and analogous standards on a case-by-case basis.

8. Key Uncertainties and Sensitive Populations

Several categories of uncertainty and population-specific susceptibility are relevant to the case-specific application of these toxicity values and should be considered whenever RoseyFields applies this methodology to an individual claimant or site:

Genetic susceptibility: Polymorphisms in benzene-metabolizing enzymes create substantial interindividual variability in susceptibility to benzene hematotoxicity. In addition to the CYP2E1/NQO1 interaction reported by Rothman et al. (1997), Seaton et al. (1994) documented up to 13-fold differences in hepatic CYP2E1 activity among individual human subjects, and Rossi et al. (1999) found that genetic polymorphisms at multiple loci (including CYP2D6, GSTT1, and NQO1) contributed to variability in urinary benzene metabolite excretion among occupationally exposed subjects. These findings indicate that some individuals may be substantially more susceptible to benzene hematotoxicity than the population-average intraspecies UF of 10 might suggest, which is a relevant consideration where a specific claimant’s genetic or metabolic profile is at issue.

Gender differences: The evidence regarding gender-based differences in susceptibility is mixed and pathway-dependent. Brown et al. (1998), applying PBPK modeling to data from Sato et al. (1975), found that women metabolize substantially more benzene than men under equivalent exposure, suggesting greater potential susceptibility in women; however, the underlying principal study (Rothman et al., 1996a) did not find a significant difference in hematologic response by gender. In experimental animals, the direction of the gender effect varies by species and endpoint: male mice are generally more sensitive to benzene-induced hematotoxicity and genotoxicity than female mice (NTP, 1986), whereas female rats have been reported to be more sensitive than male rats in the same bioassay.

Childhood and developmental susceptibility: Children may be more susceptible to benzene’s hematotoxic and leukemogenic effects than adults, both because of physiological differences (a higher inhalation rate relative to body weight, increased activity levels) and because the hematopoietic system undergoes substantial maturation and differentiation during childhood. Epidemiologic studies have reported associations between parental occupational benzene exposure (both prior to conception and during pregnancy) and increased incidence of childhood leukemia, including studies by McKinney et al. (1991), Shaw et al. (1984), Shu et al. (1988), and Buckley et al. (1989). Experimental studies in mice have similarly shown that in utero benzene exposure produces lasting alterations in hematopoietic progenitor cell populations detectable into adulthood.

Absence of a threshold at low exposure concentrations: Occupational epidemiologic evidence suggests that hematologic effects of benzene may occur at exposure concentrations well below the LOAELs identified in the principal dose-response studies, with no clearly defined threshold. Dosemeci et al. (1996) reported significantly elevated relative risk of clinical “benzene poisoning” (defined by specific WBC and platelet count criteria) in Chinese factory workers exposed to as little as 5–19 ppm compared with workers exposed to less than 5 ppm, and Bogadi-Sare et al. (2000) reported reduced circulating B-lymphocyte counts in shoe-industry workers exposed to benzene concentrations below 15 ppm. This evidence is relevant when evaluating case-specific exposure scenarios involving relatively low benzene concentrations, and it underscores the conservative, health-protective nature of the composite uncertainty factors applied in the RfC/RfD derivation.

Database completeness: As reflected in the database uncertainty factor applied throughout this methodology, the absence of a two-generation reproductive/developmental toxicity study for benzene represents an acknowledged gap in the overall toxicological database, notwithstanding the otherwise extensive body of available hematotoxicity, genotoxicity, and mechanistic literature.

Taken together, these uncertainties inform both the selection of health-protective composite uncertainty factors in the derivation of the RfC and RfD, and the case-specific judgment RoseyFields applies when characterizing risk for a particular claimant, facility, or community exposure scenario.

References