Benzo[a]pyrene & PAHs Risk Assessment and Cancer Risk

RoseyFields conducts Benzo[a]pyrene (representing PAHs) risk assessments using the following methodology: this page is based on EPA’s 2017 IRIS Toxicological Review of Benzo[a]pyrene (final report), which derived the current oral reference dose (RfD), inhalation reference concentration (RfC), oral cancer slope factor (CSF), and inhalation unit risk (IUR) for benzo[a]pyrene (BaP), and on ATSDR’s 2022 Guidance for Calculating Benzo(a)pyrene Equivalents, which provides the accepted approach for extending BaP’s cancer toxicity values to mixtures of polycyclic aromatic hydrocarbons (PAHs) encountered at contaminated sites. Citations to specific findings and numeric values below are to the underlying peer-reviewed studies rather than to the EPA or ATSDR documents themselves.

1. Chemical Identity, Regulatory Background, and Exposure Sources

Benzo[a]pyrene (BaP; CAS No. 50-32-8) is a five-ring polycyclic aromatic hydrocarbon (PAH) composed exclusively of fused aromatic (benzene) rings containing only carbon and hydrogen. PAHs as a class — more than 100 individual congeners — are formed during the incomplete combustion of coal, oil, natural gas, wood, garbage, and other organic material, including tobacco and charbroiled meat, and also occur in coal tar, coal-tar pitch, creosote, asphalt, crude oil, and roofing tar. BaP is the most extensively studied and best-characterized PAH congener, both because of the large body of available toxicological and analytical data and because of well documented, frequent human exposure to it, and it therefore serves as the benchmark compound against which the relative carcinogenic potency of other PAH congeners is expressed (see Section 8, below).

BaP and PAH mixtures are regulated and evaluated by multiple federal and state agencies. EPA has classified BaP as “carcinogenic to humans” under its 2005 Guidelines for Carcinogen Risk Assessment, a classification that supersedes the chemical’s earlier (1992) EPA Group B2 “probable human carcinogen” classification; the International Agency for Research on Cancer (IARC) has similarly classified BaP as Group 1, carcinogenic to humans. BaP and PAHs as a class are regulated under the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) as hazardous substances routinely evaluated at Superfund and other contaminated sites, and PAHs are commonly encountered contaminants at former manufactured gas plant sites, wood-treating and creosote facilities, coking and coal-tar distillation operations, aluminum smelters, asphalt and roofing operations, and sites with historical fuel or coal combustion residues.

Exposure to BaP and other PAHs occurs by all three principal routes. Oral exposure occurs primarily through ingestion of contaminated soil (particularly relevant for young children with elevated soil/dust ingestion rates and hand-to-mouth behavior) and, to a lesser extent, through consumption of PAH-contaminated food (e.g., charbroiled or smoked foods) or drinking water. Inhalation exposure occurs through breathing ambient or occupational air containing PAH-laden particulates — historically the dominant route of exposure and effect documentation in occupational cohorts such as aluminum smelter and coke-oven workers, coal-gasification workers, and chimney sweeps, and a continuing route of concern near combustion sources, in tobacco smoke, and in communities near industrial or vehicular emission sources. Dermal exposure can occur through direct soil or product contact (e.g., creosote-treated wood, coal tar, asphalt) and is a particular concern for skin cancer risk in occupational and community settings with significant dermal contact potential.

2. Point-of-Departure (POD) Selection: Critical Studies and Endpoints

EPA’s 2017 IRIS assessment derived separate points of departure for four toxicity value categories — the oral RfD, the inhalation RfC, the oral CSF, and the IUR — each grounded in a specific critical study and endpoint selected from the broader BaP toxicology database using explicit study-quality and dose-response criteria (study design, multiple exposure levels, statistical power at low doses, and relevance of exposure route/duration).

2.1 Noncancer Oral RfD: Developmental Neurotoxicity (Chen et al., 2012)

For the oral RfD, EPA evaluated candidate PODs across four hazard categories identified as human hazards or potential hazards of BaP: developmental toxicity (including developmental neurotoxicity), male reproductive toxicity, female reproductive toxicity, and immunotoxicity. Within the developmental toxicity category, EPA compared Chen et al. (2012), a well-conducted study evaluating multiple developmental and neurobehavioral endpoints in neonatal, adolescent, and adult rats following early postnatal BaP exposure (postnatal days 5–11, doses of 0.02, 0.2, and 2 mg/kg-day), against Jules et al. (2012) (in utero exposure producing cardiovascular effects in adult offspring) and several earlier, higher-dose studies (Bouayed et al., 2009; Mackenzie and Angevine, 1981; Kristensen et al., 1995) that were not carried forward because of much higher dose levels or single-dose designs that limited their utility for characterizing the low-dose region. Chen et al. (2012) reported increased locomotion in the open field test, increased latency in negative geotaxis and surface righting tests, decreased anxiety-like behavior in the elevated plus maze, and impaired performance in the Morris water maze at various developmental time points; of these, the open field activity, elevated plus maze, and Morris water maze endpoints were identified as the most compelling and were selected for benchmark dose (BMD) modeling, using a benchmark response (BMR) of a 1 standard deviation (SD) change from control. This study was ultimately selected as the basis for the overall RfD because its composite uncertainty factor was the smallest among all candidate values, it captured effects across multiple neurobehavioral domains, and similar neurobehavioral effects were replicated across numerous other studies in the wider database.

2.2 Noncancer Oral RfD: Organ/System-Specific Candidate PODs

In addition to the developmental neurotoxicity POD that formed the basis of the overall RfD, EPA derived organ/system-specific candidate values for other hazard categories, useful for cumulative or endpoint-specific risk assessment. For reproductive toxicity, Xu et al. (2010) (decreased ovary weight and primordial follicle counts in female rats following combined BaP and phthalate exposure) was selected over Gao et al. (2011) (cervical epithelial hyperplasia, a less-replicated endpoint) and over the male reproductive studies Zheng et al. (2010) (decreased intratesticular testosterone, which did not show a clear dose-response pattern) and Mohamed et al. (2010) (decreased sperm count and motility, which could not support BMD modeling because the study did not report control-group variability, resulting in a composite uncertainty factor too large — over 3,000-fold — to be used consistent with EPA’s guidance against application of the full 10-fold factor in four or more areas of extrapolation). For immunotoxicity, Kroese et al. (2001) (decreased thymus weight in a 90-day rat gavage study) and De Jong et al. (1999) (decreased serum IgM, IgA, and B-cell counts in a 35-day rat study) yielded comparable, converging candidate values and were jointly selected as most representative of BaP immunotoxicity.

2.3 Noncancer Inhalation RfC: Decreased Embryo/Fetal Survival (Archibong et al., 2002)

For the inhalation RfC, the most sensitive POD among the inhalation candidate values was decreased embryo/fetal survival (increased resorptions) in pregnant rats reported by Archibong et al. (2002), observed at a LOAEL human equivalent concentration (HEC) of 4.6 μg/m3; a NOAEL was not identified in this study. This finding is corroborated by a female-reproductive-toxicity companion study, Archibong et al. (2012), reporting decreased ovulation rate and decreased ovary weight in rats following inhalation BaP exposure, and by decreased fetal survival observed via the oral route at much higher doses in Mackenzie and Angevine (1981). Archibong et al. (2002) was selected as the principal study because it identified the lowest, most biologically significant effect level among the available inhalation studies and because the developing organism was identified as a consistently sensitive target across both oral and inhalation routes.

2.4 Oral Cancer Slope Factor: Alimentary Tract Tumors in Mice (Culp et al., 1998) and Rats (Kroese et al., 2001)

For the oral CSF, EPA identified three chronic (approximately 2-year) oral cancer bioassays in the database: a Wistar rat gavage study observing tumors at multiple sites, including the forestomach, oral cavity, liver, jejunum/duodenum, kidney, auditory canal, and skin/mammary gland (Kroese et al., 2001); a Sprague-Dawley rat dietary/gavage study observing forestomach tumors (Brune et al., 1981); and a B6C3F1 mouse dietary study observing forestomach, esophageal, tongue, and laryngeal tumors in females (Culp et al., 1998). The Kroese et al. (2001) and Culp et al. (1998) studies were selected as the best available bioassays for dose-response analysis because both were conducted under Good Laboratory Practice with adequate group sizes (approximately 50/sex/group), multiple dose levels plus controls, up to two years of treatment, and detailed histopathological and individual-animal reporting; the Brune et al. (1981) study was not selected because of its nonstandard, intermittent dosing protocol and more limited reporting.

2.5 Inhalation Unit Risk: Upper Respiratory Tract Tumors in Hamsters (Thyssen et al., 1981)

For the IUR, the only lifetime inhalation cancer bioassay available for BaP is the study by Thyssen et al. (1981) in male Syrian golden hamsters, exposed by inhalation to BaP adsorbed on sodium chloride aerosol for up to 2.5 years across three exposure concentrations (time-weighted average group concentrations of 0.25, 1.01, and 4.29 mg/m3) plus control. This study observed increasing incidences of benign and malignant tumors of the pharynx, larynx, trachea, esophagus, nasal cavity, and forestomach with increasing exposure, occurring earlier in higher-exposure groups, and included detailed individual-animal pathology and survival data. Several supporting intratracheal instillation hamster studies (e.g., Feron and Kruysse, 1978; Saffiotti et al., 1972) corroborate the carcinogenic response but were not used quantitatively because intratracheal instillation alters particle deposition, clearance, and retention relative to true inhalation exposure.

3. Pharmacokinetic Modeling and Dose Conversion (Animal-to Human, Internal-to-External Dose)

No validated chemical-specific physiologically based pharmacokinetic (PBPK) model is available for BaP, despite an extensive toxicokinetic literature, so EPA’s animal-to-human dose conversions rely on EPA’s standard default methodologies rather than a mechanistic PBPK model.

3.1 Oral Route: Body-Weight-to-the-Three-Quarters-Power (BW¾) Scaling

For the noncancer oral RfD, administered animal doses (mg/kg-day) were converted to human equivalent doses (HEDs) using BW¾ allometric scaling (i.e., multiplying the animal dose by (animal body weight/human body weight)¹⁄⁴, using a 70 kg reference human body weight), except for the developmental neurotoxicity POD from Chen et al. (2012), where dosing was administered directly to early postnatal pups; because BW¾ scaling was not considered appropriate for extrapolating doses administered directly to neonatal animals, no cross-species dosimetric adjustment was applied for that study, and the residual uncertainty was instead addressed through the interspecies uncertainty factor (Section 4). The same BW¾ approach was applied for the oral cancer slope factor: because BaP’s mutagenic mode of action is mediated by DNA-reactive metabolites formed in the target tissues, and because a chemical-specific dose metric that would more precisely capture cross-species differences in metabolic activation was not available, EPA adopted BW¾ scaling as its default, on the reasoning that overall distribution and metabolism of BaP to reactive intermediates is expected to be roughly proportional to metabolic rate, which itself scales approximately with BW¾.

3.2 Inhalation Route: Default Dosimetry versus Equal-Concentration Assumption

For the noncancer inhalation RfC, EPA applied its standard default dosimetric adjustment methodology for particulates to convert animal exposure concentrations to human equivalent concentrations (HECs), accounting for species differences in respiratory tract deposition and minute ventilation. For the inhalation unit risk, however, this same default particle-dosimetry methodology could not be applied to the Thyssen et al. (1981) hamster bioassay because that methodology was developed for insoluble, non-hygroscopic particles, whereas the Thyssen et al. (1981) study used BaP adsorbed onto a soluble, hygroscopic sodium chloride carrier aerosol. In the absence of a scientifically supportable basis for species-specific dosimetric adjustment for this exposure medium, and because relative surface areas of the upper respiratory tract in hamsters versus humans were not considered a reliable basis for extrapolation (hamsters and humans are not expected to develop tumors in anatomically equivalent respiratory tract regions), EPA instead assumed equal risk per unit air concentration (μg/m3) across species — mathematically equivalent to assuming that both breathing rate and hepatic/pulmonary metabolic activation of BaP to DNA-reactive metabolites scale with BW¾. EPA’s sensitivity analysis found that alternative scaling assumptions (BW¾ or body weight-to-the-two-thirds-power, reflecting relative surface area) would instead increase the unit risk estimate by approximately 5- to 8-fold relative to the equal-concentration default, illustrating that interspecies dose-metric selection is a material source of uncertainty in the inhalation cancer risk value (see Section 5, below).

3.3 Benchmark Dose (BMD) Modeling as the POD Method

Across all four toxicity values, EPA preferred benchmark dose (BMD) modeling over simple NOAEL/LOAEL identification where the underlying data supported it, because BMD modeling uses the full dose-response data set rather than only the tested dose levels, and it enabled derivation of a benchmark dose lower confidence limit (BMDL) as the POD — the lower 95% confidence bound on the dose associated with a defined benchmark response (BMR), e.g., a 1 SD change from control for continuous endpoints (Chen et al., 2012; Kroese et al., 2001), a 10% change from control for other continuous endpoints (Gao et al., 2011), or 10% extra risk for the dichotomous cancer bioassay tumor-incidence data (Kroese et al., 2001; Culp et al., 1998; Thyssen et al., 1981). For the cancer bioassays specifically, EPA applied the multistage-Weibull time-to-tumor model, which incorporates each animal’s time of death together with dose and tumor status, to account for competing risks and early mortality in the high-dose groups (which in the Culp et al. (1998) and Kroese et al. (2001) bioassays were largely dead or moribund well before the end of the nominal 2-year study period).

4. Uncertainty Factor Derivation

Consistent with EPA’s standard RfD/RfC methodology, up to five uncertainty factors (UFs) were considered and applied, as warranted, to each noncancer candidate POD: a LOAEL-to-NOAEL factor (UFL), a subchronic-to chronic factor (UFS), an interspecies factor (UFA), an intraspecies factor (UFH), and a database-completeness factor (UFD). The specific values applied to the RfD’s principal study (Chen et al., 2012) and to the other organ/system specific candidate PODs are summarized below.

4.1 LOAEL-to-NOAEL (UFL)

A UFL of 1 was applied wherever the POD was a NOAEL (e.g., Zheng et al., 2010; De Jong et al., 1999) or was derived via BMD modeling using a BMR judged to represent a minimal biologically significant response (a 1 SD change, as for Chen et al., 2012 and Kroese et al., 2001, or a 10% change, as for Gao et al., 2011) — BMD-derived PODs are treated as NOAEL-equivalent because the BMDL approximates a no-effect dose without relying on the specific, potentially widely spaced, dose levels actually tested. Where a true NOAEL could not be identified and the LOAEL response was of high magnitude — as in Jules et al. (2012) (15–33% increases in systolic/diastolic blood pressure at the only dose tested) and Mohamed et al. (2010) (50% and 20% decreases in sperm count and motility, respectively, at the lowest dose tested, without sufficient reporting of control variability to permit BMD modeling) — a UFL of 10 was applied to approximate the LOAEL-to-NOAEL gap.

4.2 Subchronic-to-Chronic (UFS)

A UFS of 10 was applied to PODs derived from subchronic studies (the reproductive and immunological studies in the database ranged from 42 to 90 days in duration), to account for the possibility that longer exposure durations could produce effects at lower doses. A UFS of 1 was applied to the developmental toxicity PODs (Chen et al., 2012; Jules et al., 2012), because dosing occurred during gestation or the early postnatal period specifically to capture a discrete, biologically defined developmental exposure window, for which EPA’s guidance recognizes that exposure duration adjustment to a full lifetime is not scientifically warranted — the developmental period itself, not cumulative lifetime exposure, is the relevant susceptibility window for these effects.

4.3 Interspecies (UFA)

An interspecies UFA of 3 (10¹⁄² = 3.16, rounded to 3) was applied to all noncancer oral candidate PODs except the Chen et al. (2012) developmental neurotoxicity POD, reflecting EPA’s standard practice of applying a reduced (rather than full 10-fold) interspecies factor when BW¾ scaling has already been used to address the toxicokinetic component of cross-species extrapolation, leaving only the toxicodynamic component (a factor of √10 ≈ 3) as residual uncertainty. For Chen et al. (2012), because BW¾ scaling was not applied (doses were administered directly to neonatal pups rather than derived via allometric conversion; see Section 3.1), a full interspecies UFA of 10 was applied instead, to account for the complete absence of quantitative information addressing either toxicokinetic or toxicodynamic cross-species differences at this early postnatal life stage.

4.4 Intraspecies/Human Variability (UFH)

A full intraspecies UFH of 10 was applied to every noncancer candidate POD, reflecting the use of inbred laboratory animal strains that are not considered representative of the toxicokinetic and toxicodynamic variability present in the general human population, including sensitive subgroups such as the developing fetus. Because no PBPK model or other chemical-specific human variability data were available for BaP to support replacing any portion of this default factor, the full UFH of 10 was retained across all candidate values, for both the RfD and RfC.

4.5 Database Completeness (UFD)

For the oral RfD, a UFD of 3 was applied to the selected developmental-neurotoxicity POD (Chen et al., 2012) to account for the absence of a standard multigenerational or extended one-generation reproductive/developmental toxicity study spanning premating through lactation — a design that could plausibly identify a more sensitive POD than the single-window postnatal exposure study used — but a full UFD of 10 was judged unwarranted because the selected POD already represented some of the most sensitive effect levels in the entire BaP database, including relative to other low-dose developmental studies, and was corroborated by numerous replicate studies. For the inhalation RfC, by contrast, a full UFD of 10 was applied, reflecting the substantially thinner inhalation database relative to the oral database: the inhalation database lacks a multigenerational study, lacks any developmental neurotoxicity study (despite oral-route and PAH-mixture evidence that the developing nervous system is a sensitive target), and lacks studies characterizing subchronic or chronic immune system effects following inhalation exposure.

4.6 Composite Uncertainty Factors Applied

• Overall oral RfD (Chen et al., 2012, neurobehavioral changes): UFL = 1, UFS = 1, UFA = 10, UFH = 10, UFD = 3 — composite UF = 300.
• Reproductive toxicity oral candidate value (Xu et al., 2010, ovotoxicity): UFL = 1, UFS = 10, UFA = 3, UFH = 10, UFD = 3 — composite UF = 1,000.
• Immunotoxicity oral candidate value (Kroese et al., 2001 / De Jong et al., 1999): UFL = 1, UFS = 10, UFA = 3, UFH = 10, UFD = 3 — composite UF = 1,000.
• Overall inhalation RfC (Archibong et al., 2002, decreased embryo/fetal survival): UFL = 10, UFS = 1, UFA = 3, UFH = 10, UFD = 10 — composite UF = 3,000.

No uncertainty factors are applied in the derivation of the cancer slope factor or inhalation unit risk; instead, uncertainty in those values is addressed through the choice of a linear low-dose extrapolation approach and the use of the 95% lower confidence bound on the benchmark dose/concentration (BMDL/BMCL) as the point of departure for that extrapolation, which is discussed further in Section 6.

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

5.1 Organ/System-Specific and Overall Oral RfD

EPA derived organ/system-specific RfDs — useful for cumulative risk assessments considering multiple agents acting at a common target — for developmental toxicity (3 × 10 ⁴ mg/kg-day, based on Chen et al. (2012) ⁻ neurobehavioral changes; medium confidence), reproductive toxicity (4 × 10 ⁴ mg/kg-day, based on Xu et al. ⁻ (2010) ovotoxicity; medium confidence), and immunotoxicity (2 × 10 ³ mg/kg-day, based on Kroese et al. ⁻ (2001)/De Jong et al. (1999) decreased thymus weight and serum IgM; low confidence). The overall RfD was selected as the lowest of these organ/system-specific values with the highest associated confidence: the overall oral RfD for benzo[a]pyrene is 3 × 10 ⁴ mg/kg-day (i.e., 0.0003 mg/kg-day), based on developmental neurobehavioral ⁻ changes in rats exposed during early postnatal life (Chen et al., 2012), with an assigned overall confidence of medium. This value is intended to be protective of the general population, including sensitive subgroups, for a lifetime of continuous exposure, though EPA specifically cautioned that because the critical effect derives from a discrete early-life exposure window (postnatal days 5–11 in rats, believed to correspond developmentally to the human third trimester), time-averaging exposure over a longer period for comparison with the RfD may understate risk if exposure is disproportionately elevated during early development.

5.2 Organ/System-Specific and Overall Inhalation RfC

For the inhalation route, an organ/system-specific RfC for reproductive toxicity was derived from decreased ovulation rate in Archibong et al. (2012); a candidate value for male reproductive toxicity (decreased testis weight) could not be derived because its composite uncertainty factor exceeded 30,000-fold. The overall inhalation RfC for benzo[a]pyrene is 2 × 10 ⁶ mg/m3, based on decreased embryo/fetal survival (increased resorptions) in rats ⁻ following inhalation exposure during gestation (Archibong et al., 2002), with an assigned overall confidence of low to-medium (medium confidence in the principal study; low confidence in the database, reflecting the same database gaps discussed in Section 4.5). As with the RfD, the RfC is intended to protect the general population, including sensitive subgroups, and EPA specifically identified the first (and potentially second) trimester of pregnancy as the likely window of susceptibility for the critical effect, again cautioning that time-averaged exposure comparisons may not be protective if exposure is elevated specifically during this window.

6. Carcinogenicity Classification and Derivation of the Cancer Slope Factor (CSF) and Inhalation Unit Risk (IUR)

6.1 Weight-of-Evidence Classification: “Carcinogenic to Humans”

Under EPA’s 2005 Guidelines for Carcinogen Risk Assessment, benzo[a]pyrene is classified as “carcinogenic to humans.” This descriptor requires: (a) strong evidence of an association between human exposure and cancer or the key precursor events of the carcinogenic mode of action; (b) extensive evidence of carcinogenicity in animals; (c) identification of the carcinogenic mode(s) of action and associated key precursor events in animals; and (d) strong evidence that those key precursor events are anticipated to occur in humans and progress to tumors. For BaP, human evidence derives substantially from occupational and environmental exposure to complex PAH mixtures (coal tar, coal-tar pitch, soot, tobacco smoke) rather than from BaP exposure in isolation; nonetheless, methodologically strong occupational cohort studies with BaP-specific (rather than only total-PAH) exposure metrics provide consistent evidence of an association with lung cancer, with approximately a 2-fold increased risk at higher cumulative BaP exposures (measured in μg/m3-years) after adjustment for smoking, replicated across three large studies in different countries and industries (Armstrong and Gibbs, 2009; Spinelli et al., 2006; Xu et al., 1996). A comparable, though somewhat less consistent, pattern of increasing bladder cancer risk with increasing cumulative BaP exposure (above roughly 80 μg/m3-years) was observed in two of three aluminum-industry cohort studies with detailed BaP exposure data (Gibbs and Sevigny, 2007a, 2007b; Gibbs et al., 2007; Spinelli et al., 2006). Mechanistically, elevated benzo[a]pyrene-7,8-diol-9,10-epoxide (BPDE)-DNA adduct levels have been documented in white blood cells of coke-oven workers and chimney sweeps and in lung tissue of smokers with lung cancer, and mutation spectra characteristic of diol-epoxide-induced damage (G→T transversions) have been identified in the p53 and K-ras genes of lung tumors from individuals with heavy PAH-mixture exposure, providing biological plausibility and coherence between the animal mode-of-action data and the human epidemiological findings.

6.2 Mutagenic Mode of Action and Linear Low-Dose Extrapolation

EPA concluded that BaP carcinogenicity involves a mutagenic mode of action, mediated by DNA-reactive metabolites (principally BPDE) formed through cytochrome P450 and epoxide hydrolase-mediated metabolism, which form covalent DNA adducts and induce characteristic mutations in tumor-relevant genes. Consistent with EPA’s Guidelines for Carcinogen Risk Assessment, a linear low-dose extrapolation approach was therefore used for both the oral CSF and the IUR, rather than a nonlinear (threshold) approach.

6.3 Oral Slope Factor: 1 per mg/kg-day

Using the multistage-Weibull time-to-tumor model, PODs (BMDL10, the 95% lower bound on the dose associated with 10% extra tumor risk) were estimated separately by sex and tumor site for the Kroese et al. (2001) rat bioassay and the Culp et al. (1998) mouse bioassay. Human-equivalent slope factors (calculated as 0.1/BMDL10,HED) ranged from 0.04 per mg/kg-day (male rat kidney tumors) to 1 per mg/kg-day (female mouse alimentary tract tumors), with liver tumor-based slope factors in rats (0.2 per mg/kg-day) close to the alimentary-tract-tumor-based values (0.2–0.3 per mg/kg-day) in the same species. A composite slope factor combining all tumor sites in male rats was approximately 0.5 per mg/kg-day, only about 25% higher than the single most sensitive rat tumor site; the geometric mean across all three data sets (male rats, female rats, female mice) was approximately 0.6 per mg/kg day, about 40% of the highest single-study value. Because EPA’s cancer guidelines direct that the most representative single data set be selected where scientifically justified, and because no basis existed to prefer the rat or mouse results as more representative of human response, EPA selected the most sensitive (health-protective) result: the oral slope factor for benzo[a]pyrene is 1 per mg/kg-day, based on the alimentary tract tumor response (forestomach, esophagus, tongue, larynx) in female B6C3F1 mice (Culp et al., 1998). This value is intended for use only at lifetime average daily doses below approximately 0.1 mg/kg-day, above which the dose-response relationship becomes nonlinear as risk approaches saturation.

6.4 Inhalation Unit Risk: 6 × 10 ⁴ per μg/m3 ⁻

Using the same multistage-Weibull modeling approach applied to the Thyssen et al. (1981) hamster inhalation bioassay (combined benign and malignant tumors of the pharynx, larynx, trachea, and nasal cavity), and treating tumors as incidental (rather than causal) to death — the more appropriate assumption given that cancer risk assessment focuses on incidence rather than tumor-caused mortality — EPA derived a BMCL10 of 0.163 mg/m3. Linear extrapolation from this BMCL10 (slope = 0.1/BMCL10) yields an inhalation unit risk of 0.6 per mg/m3, equivalent to 6 × 10 ⁴ per μg/m3. This value is intended for use only at lifetime average air concentrations below ⁻ approximately 0.3 mg/m3 (the human-equivalent POD), above which the dose-response relationship becomes nonlinear.

6.5 Age-Dependent Adjustment Factors (ADAFs) for Early-Life Exposure

Because BaP’s carcinogenicity operates via a mutagenic mode of action, and because the CSF and IUR were derived from adult-only animal exposures that do not reflect the generally increased susceptibility of early life stages to mutagenic carcinogens, EPA’s Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens requires application of age-dependent adjustment factors (ADAFs) when estimating cancer risk from exposures occurring before age 16: an ADAF of 10 for exposures occurring at less than 2 years of age, an ADAF of 3 for exposures occurring between 2 and less than 16 years of age, and an ADAF of 1 (i.e., no adjustment) for exposures at or after age 16. These ADAFs are applied by multiplying the unit risk (CSF or IUR) by the applicable ADAF for each age-specific exposure period, then summing the age-specific risk contributions across the full exposure duration being evaluated; EPA’s own worked example, for a hypothetical constant lifetime oral exposure of 0.001 mg/kg-day, shows that appropriately weighting the 0–<2-year, 2–<16-year, and ≥ 16-year exposure periods by their respective ADAFs and duration fractions yields a total 70-year cancer risk of 2 × 10 ³ — meaningfully ⁻ higher than a risk estimate that applied the adult (ADAF = 1) unit risk uniformly across the full lifetime.

7. Application in Practice: Hazard Quotients, Cancer Risk, and PAH Mixture Evaluation

7.1 Noncancer Screening: Hazard Quotient and Hazard Index

For noncancer effects, RoseyFields’ risk assessors calculate a hazard quotient (HQ) for each relevant exposure pathway by dividing the estimated chronic daily intake or air concentration (the exposure point concentration, or EPC, appropriately adjusted for exposure frequency, duration, and body weight) by the applicable toxicity value — the oral RfD (3 × 10 ⁴ mg/kg-day) for ingestion/dermal pathways or the inhalation RfC (2 × 10 ⁶ mg/m3) for ⁻ ⁻ inhalation pathways. An HQ at or below 1 indicates that estimated exposure is below the level associated with appreciable risk of adverse noncancer effects; an HQ above 1 indicates a potential noncancer hazard warranting further evaluation. Where multiple PAH congeners or other co-occurring contaminants share a common toxicological target or mechanism, hazard quotients for each are typically summed into a hazard index (HI) for that target organ/system — which is one reason EPA’s organ/system-specific RfDs and RfCs (developmental, reproductive, immunological) are particularly useful, since they allow congener- and pathway-specific HQs to be summed for the specific organ system of concern rather than only against the single overall RfD/RfC. Consistent with the ATSDR guidance, however, BaP-equivalent concentrations (Section 7.3, below) are never used for noncancer hazard quotient calculations; each PAH congener with an available noncancer toxicity value is evaluated individually against its own congener-specific RfD, RfC, or other health-based comparison value.

7.2 Cancer Risk: Excess Lifetime Cancer Risk

For carcinogenic effects, RoseyFields calculates excess lifetime cancer risk (ELCR) by multiplying the estimated lifetime average daily dose (LADD, in mg/kg-day) by the oral CSF (1 per mg/kg-day) for ingestion/dermal exposure, or the estimated lifetime average air concentration (in μg/m3) by the IUR (6 × 10 ⁴ per μg/m3) for ⁻ inhalation exposure. Where exposure occurs during childhood or adolescence, age-dependent adjustment factors (Section 6.5) are applied to each relevant age bracket before summing age-specific risk contributions, rather than applying the adult-derived CSF or IUR uniformly across the full exposure duration — omitting this step can materially understate risk for early-life exposure scenarios, which are common in residential and school/daycare exposure settings evaluated in public health assessments and toxic-tort matters. Resulting ELCR estimates are then compared against the risk range typically used in regulatory and public health decision-making (often 1 × 10 ⁶ to 1 ⁻ × 10 ⁴, consistent with EPA Superfund practice), recognizing that both the CSF and IUR represent upper-bound ⁻ (not central-tendency) estimates of risk, such that the true risk at a given exposure level is expected to be at or below the calculated ELCR.

7.3 Evaluating PAH Mixtures: BaP-Equivalent Concentrations

Because BaP rarely occurs in isolation at contaminated sites — PAHs are near-universally encountered as complex mixtures reflecting their shared combustion or petrogenic origin — RoseyFields follows ATSDR’s 2022 potency equivalency-factor (PEF) methodology, itself based on the California Office of Environmental Health Hazard Assessment’s (OEHHA) PEF derivations (Collins et al., 1998), to evaluate PAH cancer risk as a class. Each measured PAH congener’s concentration is multiplied by its congener-specific PEF (expressing that congener’s carcinogenic potency relative to BaP, whose PEF is defined as 1) to yield a congener-specific BaP-equivalent concentration (BEC); the BECs for all measured congeners in a sample are then summed to produce a total BaP equivalent for that sample. PEFs for the class range from 0.01 (e.g., chrysene) to 10 (e.g., dibenzo[a,h]pyrene, dibenzo[a,i]pyrene, dibenzo[a,l]pyrene), reflecting up to a 1,000-fold range in relative carcinogenic potency across PAH congeners with sufficient toxicological data to support a PEF. RoseyFields’ standard workflow follows ATSDR’s four-step approach: (1) calculate a BaP equivalent for each environmental sample and screen the maximum sample-level BaP equivalent against the applicable cancer comparison value for BaP; (2) if the screening step is exceeded, calculate an exposure point concentration (EPC, typically a 95% upper confidence limit on the mean) for each individual PAH congener across all samples in the exposure unit; (3) apply congener-specific PEFs to each congener EPC and sum the results into a BaP-equivalent EPC; and (4) calculate cancer risk using the BaP equivalent EPC together with the oral CSF (or IUR, for the inhalation pathway) for BaP itself, since the BaP equivalent value expresses the entire measured PAH mixture’s carcinogenicity in BaP-normalized units. Naphthalene is specifically excluded from BaP-equivalent calculations — unlike the other PAH congeners with

PEFs, naphthalene is not mutagenic, and its evidence for oral carcinogenicity is limited, so it is evaluated separately with its own congener-specific toxicity values (and only for the inhalation pathway, given the limited oral carcinogenicity evidence). Non-detect PAH results are addressed conservatively: during initial screening, non detects are set to the full detection limit; where every sample for a given congener is a non-detect, precluding calculation of a standard EPC, an upper-bound/lower-bound sensitivity analysis is performed (setting the congener’s EPC alternately to the maximum detection limit and to zero) to characterize the influence of that congener’s uncertain concentration on the overall cancer risk estimate.

7.4 Application in Litigation and Regulatory Support

In litigation and regulatory support work, these toxicity values and calculation methods serve several related functions. First, HQ/HI and ELCR calculations, built on dose or exposure reconstruction specific to a site or individual (source characterization, environmental fate and transport modeling, and exposure-pathway-specific intake assumptions for the population and time period at issue), provide the quantitative backbone for general causation opinions regarding whether the chemical is capable of causing the alleged health effect at the exposure levels documented or reconstructed for the case. Second, because the RfD, RfC, CSF, and IUR each derive from an identified critical study and explicit, disclosed uncertainty-factor or extrapolation methodology (Sections 2–6), RoseyFields’ risk assessors can transparently explain and, where necessary, defend each element of a risk calculation under expert-testimony scrutiny, including identifying which specific study and endpoint drove a given conclusion and why alternative studies or endpoints were not selected. Third, because BaP’s toxicity values are explicitly designed to be paired with PEFs for whole-PAH-mixture risk assessment (Section 7.3), RoseyFields can extend a single, well-characterized toxicological foundation to the full suite of PAH congeners typically encountered together at a contaminated site, rather than needing (or having available) a complete independent toxicological derivation for every individual congener. Finally, because ADAFs are required for any cancer risk estimate involving childhood or adolescent exposure, RoseyFields’ litigation exposure-reconstruction work routinely incorporates age-stratified exposure and risk calculations wherever a plaintiff’s or class’s exposure history includes any period before age 16, since omission of ADAFs is both scientifically unsupportable under current EPA methodology and likely to understate risk for the specific population at issue in early-life exposure cases.

8. Key Uncertainties and Sensitive Populations

8.1 Uncertainties in the Noncancer Values (RfD/RfC)

• Both the RfD and RfC are grounded in a single principal animal study (Chen et al., 2012 for the RfD; Archibong et al., 2002 for the RfC); while each is corroborated by other studies in the broader database, neither toxicity value rests on a large, independent multi-study weight of evidence in the way the cancer values do.
• The developmental neurotoxicity mode of action underlying the RfD is not fully characterized, so the precise human developmental window of greatest susceptibility, and whether the rat postnatal day 5–11 window used in Chen et al. (2012) truly corresponds to the human third-trimester analog assumed by EPA, cannot be established with current data.
• The inhalation database is materially thinner than the oral database (reflected in the full 10-fold UFD applied to the RfC), including a complete absence of inhalation developmental neurotoxicity and immunotoxicity studies — a significant gap given that both endpoints are established hazards by the oral route.
• Neither the RfD nor the RfC benefits from a validated PBPK model, so the intraspecies (UFH) and, for the RfD’s principal study, interspecies (UFA) uncertainty factors are applied at their full default values rather than being informed or reduced by BaP-specific human variability or cross-species data.

8.2 Uncertainties in the Cancer Values (CSF/IUR)

• The IUR rests on a single inhalation bioassay, in one sex, of one species (male hamsters; Thyssen et al., 1981), using a sodium chloride carrier aerosol that differs materially from the carbonaceous particulates (soot, diesel exhaust) that typically carry BaP in real-world environmental and occupational exposure, introducing uncertainty in both the dose-response shape and the appropriate interspecies dose metric (Section 3.2).
• Interspecies extrapolation for the IUR assumes equal risk at equal air concentration in the absence of chemical-specific data; EPA’s own sensitivity analysis found that alternative, arguably equally defensible allometric scaling assumptions would increase the IUR by 5- to 8-fold, meaning the selected IUR should be understood as one point within a materially wider plausible range rather than a precisely bounded estimate.
• The oral CSF is a composite judgment across three tumor data sets (male rats, female rats, female mice) that varied roughly 5-fold in their site-specific slope factor estimates; EPA selected the single highest (most health protective) value rather than a central-tendency composite, a choice that is protective by design but means the selected CSF does not represent a best estimate of expected risk.
• Both the CSF and IUR are explicitly bounded to low-dose, sub-saturation exposure ranges (below approximately 0.1 mg/kg-day for the CSF and 0.3 mg/m3 for the IUR); linear extrapolation using these values for exposure estimates above these thresholds is not scientifically supported and requires consultation of the full nonlinear dose-response model instead.

8.3 Sensitive Populations and Life-Stage Considerations

Both the noncancer and cancer toxicity values for BaP place particular emphasis on developmental and early-life exposure. The overall RfD and RfC are each grounded in developmental toxicity endpoints observed after prenatal or early postnatal exposure, reflecting the developing fetus and neonate/infant as the most sensitive subpopulation identified in the current database for noncancer effects. For cancer risk, the mutagenic mode of action and the mandatory application of ADAFs for any exposure occurring before age 16 (Section 6.5) reflect the broader, chemical-class-independent principle that mutagenic carcinogens carry disproportionately elevated risk per unit dose when exposure occurs during early life, when cell proliferation rates are higher and there is more remaining lifespan for a mutational event to progress to a clinically apparent tumor. Practically, this means that case-specific risk assessments involving any plaintiff or population with in utero, infant, childhood, or adolescent exposure require both the developmental-toxicity-based noncancer values and age-stratified ADAF-adjusted cancer risk calculations, and cannot rely on adult-exposure-only toxicity values or risk equations without materially understating the risk applicable to that population.

References