Benzo[a]pyrene & PAHs Epidemiology and Cancer Risk

RoseyFields evaluates the human epidemiological evidence for BaP as follows: this summary is based on the human evidence synthesized in EPA’s 2017 IRIS Toxicological Review of Benzo[a]pyrene (final report), and on the underlying peer-reviewed epidemiological studies identified and evaluated within that review. Because benzo[a]pyrene (BaP) essentially never occurs in isolation in real-world human exposure — it is a constituent of complex polycyclic aromatic hydrocarbon (PAH) mixtures from tobacco smoke, combustion emissions, coal tar, and related sources — nearly all of the epidemiological literature discussed below reflects populations exposed to PAH mixtures in which BaP is present and, in the strongest studies, specifically measured as an exposure metric. Citations below are to the original peer-reviewed epidemiological publications rather than to the EPA document itself.

1. Study Populations and Data Sources

The human evidence base for BaP draws on several distinct categories of study populations, each measuring BaP or BaP-related biomarkers with a different degree of exposure specificity, and together spanning a wide range of exposure levels, exposure routes, and study designs.

1.1 Occupational Cohorts

Occupational cohorts with quantitative, BaP-specific exposure assessment provide the strongest evidence base for BaP’s carcinogenic effects, particularly for lung and bladder cancer. These include cohorts of aluminum reduction plant (smelter) workers in Canada, exposed to coal-tar-pitch volatiles containing BaP (Armstrong and Gibbs, 2009; Spinelli et al., 2006; Gibbs and Sevigny, 2007a, 2007b; Gibbs et al., 2007); iron and steel workers in China (Xu et al., 1996); asphalt (road paving and roofing) workers, evaluated for both lung cancer and ischemic heart disease (Burstyn et al., 2005, 2007); and coke-oven workers, evaluated primarily for reproductive, immune, and hematological endpoints associated with PAH-mixture exposure (Hsu et al., 2006; Wu et al., 2003b; Szczeklik et al., 1994). EPA’s review specifically tiered these occupational studies by exposure-assessment quality, reserving “Tier 1” status for studies with a direct, cumulative BaP exposure metric (typically μg/m3-years), sample sizes sufficient to detect an effect (generally more than approximately 50 exposed cases for cancer endpoints), and follow-up periods adequate to capture the relevant latency (e.g., more than 20 years for lung cancer) — a substantially more rigorous standard than simply having “some PAH exposure history” in the cohort.

1.2 Community and Contamination-Related Populations

Populations with dermal or environmental exposure to coal tar, creosote, or related PAH-bearing materials provide a second line of evidence, particularly for skin cancer. These include Scandinavian registry-based studies of creosote exposed timber workers and brick makers (Karlehagen et al., 1992; Törnqvist et al., 1986), asphalt roofers (Hammond et al., 1976), and clinical populations of dermatology patients treated therapeutically with topical coal tar formulations for psoriasis and eczema (e.g., Roelofzen et al., 2010; Stern et al., 1998; Maier et al., 1996) — the latter offering a comparatively well-characterized, if imperfect, human dermal BaP exposure model.

1.3 General-Population Birth Cohorts and Biomarker Studies

For developmental, reproductive, and neurodevelopmental outcomes, the strongest evidence derives from general population birth cohort studies that measure BaP-specific biomarkers — principally benzo[a]pyrene-7,8-diol-9,10- epoxide (BPDE)-DNA adducts in maternal or umbilical cord blood — rather than relying on total-PAH or questionnaire-based exposure proxies. Two major, independently conducted birth cohorts using this approach are a New York City cohort led by Columbia University researchers (Perera et al., 2004, 2005a, 2005b, 2009, 2012a, 2012b) and a Chinese cohort based in Tongliang and other communities (Tang et al., 2006, 2008). Complementary

European cohorts in Spain and Norway used a dietary benzo[a]pyrene intake metric derived from detailed food frequency questionnaires rather than a blood biomarker (Duarte-Salles et al., 2012, 2013), providing an independent, orally-focused line of evidence that converges with the adduct-based cohorts. Smaller studies of reproductive-clinic and general-population samples (e.g., Neal et al., 2008; Gaspari et al., 2003; Xia et al., 2009; Rubes et al., 2010; Hsu et al., 2006) provide biomarker-based evidence (BaP levels in follicular fluid or serum, BPDE-DNA adducts in sperm, urinary PAH metabolites) relevant to male and female fertility.

1.4 Why Multiple, Diverse Populations Strengthen Causal Inference

Because virtually no human population is exposed to BaP alone, the central causal-inference challenge for this chemical is distinguishing BaP-attributable effects from the effects of co-occurring PAHs, other combustion products, and, in many studies, tobacco smoke itself. Multiple, independent populations spanning distinct exposure contexts — occupational inhalation (smelter, coke-oven, and asphalt workers), dermal contact (creosote and coal tar-exposed workers and dermatology patients), and general-population dietary or ambient exposure (birth cohorts using adduct or dietary-intake metrics) — substantially strengthen causal inference for two related reasons. First, each population has a different mixture composition and different co-exposure profile, so an association that replicates across occupational, dermal, and general-population contexts is unlikely to be explained by any single shared confounder or co-exposure specific to one setting. Second, studies that use a direct, BaP-specific exposure metric (cumulative occupational BaP exposure in μg/m3-years; BPDE-DNA adduct levels in blood; BaP levels in follicular fluid or dietary intake) rather than a generic “PAH mixture” or occupational-title proxy allow the exposure-response relationship to be anchored specifically to BaP, even though the biological exposure realistically remains a mixture. EPA’s evidence synthesis explicitly relied on this distinction, tiering occupational cancer studies by whether they used a direct BaP exposure measure, and weighting the developmental and reproductive literature toward studies using BPDE-DNA adducts or comparably specific biomarkers over generic PAH or smoking-status proxies.

2. Overview of Health Outcomes Investigated, Organized by Body System

EPA’s 2017 human hazard evaluation for BaP organized the epidemiological (and supporting animal) evidence into developmental toxicity (including developmental neurotoxicity), reproductive toxicity, immunotoxicity, “other” noncancer toxicity (liver, kidney, cardiovascular, adult neurotoxicity), and carcinogenicity. The table below summarizes the outcomes investigated in humans within each category, organized by evidence strength.

2.1 Cancer

• Lung cancer — STRONG, quantitatively modeled evidence (see Section 3.1); consistent, statistically controlled exposure-response relationships across three geographically and industrially independent Tier 1 occupational cohorts.
• Bladder cancer — SUGGESTIVE evidence with an exposure-dependent pattern in the highest-quality studies but overall less consistent than lung cancer; an exposure-response relationship was observed in aluminum smelter cohorts spanning a wide exposure range, but not in an asphalt-worker cohort whose exposure range was much narrower and lower, and smaller Tier 2 studies showed a mixed, largely non-statistically-significant pattern of relative risks (see Section 3.2).
• Non-melanoma skin cancer (squamous cell carcinoma) — SUGGESTIVE evidence, historically well documented (dating to 18th-century observations in chimney sweeps) but based on a smaller and more heterogeneous modern literature (creosote-exposed timber workers and brick makers, asphalt roofers, and coal tar-treated dermatology patients), with likely underestimation of risk due to early excision of precancerous lesions and low case-fatality (see Section 3.3).
• Melanoma — NULL evidence; the two occupational studies with melanoma-specific mortality data found no increased risk.

2.2 Developmental / Reproductive (Fetal and Perinatal)

• Reduced birth weight and/or head circumference — STRONG, consistent evidence across independent birth cohorts using BPDE-DNA adduct biomarkers (New York City; China) and independent European cohorts using dietary intake metrics (Spain; Norway) (see Section 3.4).
• Reduced postnatal growth (weight at 18–30 months) — STRONG/consistent within the Chinese cohort specifically, corroborating the birth-weight findings (Tang et al., 2006).
• Fetal death / miscarriage — SUGGESTIVE evidence from a single Chinese case-control study showing a strong (4-fold) association with maternal blood adduct levels (Wu et al., 2010); not independently replicated.
• Birth length — SUGGESTIVE/INCONSISTENT; positive associations in the European dietary-intake cohorts (Duarte-Salles et al., 2012, 2013) were not replicated in the U.S. or Chinese adduct-based cohorts, which found no association with birth length.

2.3 Neurodevelopmental

• Cognitive/developmental quotient deficits — STRONG, consistent evidence across independent cohorts and exposure metrics (cord blood adducts in China; ambient air PAH monitoring in New York City) (see Section 3.5).
• Neuromuscular/reflex function deficits — SUGGESTIVE, based on the Chinese cohort (Tang et al., 2008).
• Behavioral outcomes (attention problems, anxious/depressed behavior) — SUGGESTIVE, based on a single New York City cohort follow-up at age 6–7 (Perera et al., 2012b); not yet independently replicated with a comparable design.

2.4 Reproductive (Adult Fertility)

• Male sperm quality/DNA damage (oligospermia, abnormal morphology, BPDE-DNA adducts, sperm chromatin fragmentation) — SUGGESTIVE, fairly consistent across several small-to-moderate occupational and clinical studies (coke-oven workers, smokers, fertility-clinic patients), but confounded by concurrent smoking status and mixture co-exposure in most studies, and using biomarkers (e.g., sperm chromatin structure assay) whose predictive relationship to actual fertility remains scientifically uncertain (see Section 3.6).
• Female fertility / ovarian toxicity — NULL/INADEQUATE for BaP specifically; EPA’s review explicitly states that no human epidemiological studies directly relating ovotoxicity to BaP exposure were available; the only directly relevant human study (in vitro fertilization patients) examined conception rates rather than ovarian pathology (Neal et al., 2008), and broader female-fertility evidence derives from cigarette-smoking epidemiology rather than BaP-specific studies.

2.5 Immune System

• Allergic sensitization following in utero exposure — NULL; the single available BaP-specific human developmental immunotoxicity study found no association (Jedrychowski et al., 2011).
• Immunoglobulin alterations (serum IgM, IgA, IgG, IgE) — SUGGESTIVE BUT INCONSISTENT; occupational (coke-oven worker) studies report both immunosuppressive (reduced IgM/IgA) and immunostimulatory (elevated IgG or IgE) findings, and none isolate BaP from the broader PAH mixture (Wu et al., 2003b; Szczeklik et al., 1994; Karakaya et al., 1999).
• No other human immune endpoints (e.g., cellular immune function, lymphocyte subsets) have been directly evaluated in relation to BaP or PAH-mixture exposure with adequate specificity; this is a data gap rather than a null finding.

2.6 Cardiovascular

• Ischemic heart disease (IHD) incidence/mortality — SUGGESTIVE, based on two occupational cohorts (aluminum smelter workers; asphalt workers) with cumulative BaP exposure metrics; one showed a non-

statistically-significant trend with lagged exposure, and the other showed a statistically significant trend in average (but not cumulative) exposure, with an approximately 60% risk increase between lowest and highest exposure groups (see Section 3.7).

• General cardiovascular mortality — INCONSISTENT/CONFOUNDED; elevated cardiovascular mortality was observed in a coke-oven-worker cohort, but comparably elevated mortality was also observed in the cohort’s non-exposed or minimally exposed comparison population, undermining attribution to BaP or PAH exposure specifically (Chau et al., 1993).

2.7 Other Organ Systems (Liver, Kidney, Adult Neurotoxicity)

• Liver, kidney, and adult (non-developmental) neurological effects — INADEQUATE human evidence; EPA’s review identified essentially no epidemiological data specific to BaP for these endpoints and drew no conclusions regarding human hazard, relying instead on animal data alone, which itself was characterized as less robust and consistent than the developmental, reproductive, and immune evidence.

3. Key Studies, Exposure Metrics, and Effect Sizes for Priority Outcomes

3.1 Lung Cancer (Strong Evidence)

The strongest human cancer evidence for BaP concerns lung cancer. Three independent occupational cohort studies, in different countries and different industries, each used a cumulative BaP exposure metric (μg/m3-years) and each found increasing lung cancer risk with increasing cumulative exposure after statistical adjustment for smoking: a Canadian aluminum-smelter cohort (Armstrong and Gibbs, 2009), a second Canadian aluminum-smelter cohort (Spinelli et al., 2006), and a Chinese iron-and-steel-worker case-control study (Xu et al., 1996). A broader set of eight additional (“Tier 2”) occupational studies with less precise exposure assessment showed a generally consistent, though more variable, pattern: standardized mortality ratios (SMRs) ranged from 1.2 to 2.9 in eight of nine studies, with only one estimate falling below 1.0. Because all identified studies involve co-exposure to other PAHs (and, in several cohorts, other combustion products), the relative contribution of BaP specifically versus other PAH-mixture constituents cannot be fully isolated from these data; nonetheless, the consistency of a BaP-specific exposure response pattern across three methodologically strong, geographically and industrially independent cohorts provides considerable support for a causal association between BaP exposure and lung cancer.

3.2 Bladder Cancer (Suggestive Evidence)

Bladder cancer evidence is more limited by smaller case counts (which constrain the ability to characterize an exposure-response relationship) but follows a broadly consistent pattern. Three Tier 1 cohort studies with detailed BaP exposure data identified 48–90 bladder cancer cases: two aluminum-smelter cohorts (Gibbs and Sevigny, 2007a, 2007b; Gibbs et al., 2007; Spinelli et al., 2006), which found an exposure-response relationship across a wide cumulative exposure range (above approximately 80 μg/m3-years), and an asphalt-worker cohort (Burstyn et al., 2007), which found no relationship with cumulative exposure up to approximately 2 μg/m3-years — a range corresponding only to the lowest exposure categories examined in the smelter cohorts, meaning the two sets of findings are not necessarily inconsistent so much as reflective of different, non-overlapping exposure ranges. Five additional, smaller occupational studies with more limited exposure characterization found relative risks ranging from 0.6 to 2.9, none individually statistically significant, reflecting the imprecision expected from small case counts (2–16 cases per study) rather than a genuinely null overall pattern.

3.3 Non-Melanoma Skin Cancer (Suggestive Evidence)

Skin cancer, specifically squamous cell carcinoma associated with dermal PAH exposure, has the longest historical pedigree of any BaP-associated cancer, dating to Percivall Pott’s 18th-century description of scrotal cancer in chimney sweeps. Modern evidence is more mixed: recent Nordic studies of chimney sweeps have not found elevated non-melanoma skin cancer incidence, plausibly reflecting substantially improved occupational hygiene since Pott’s era; however, an asphalt-roofer cohort found an SMR of 4.0 (95% CI: 1.0, 10.9) for workers employed 20 or more years (Hammond et al., 1976), and two Scandinavian population-registry studies of creosote-exposed timber workers and brick makers found standardized incidence ratios of 1.5 (95% CI: 0.7, 2.6; 5 cases) and 2.37 (95% CI: 1.08, 4.50; 9 cases), respectively (Törnqvist et al., 1986; Karlehagen et al., 1992). Because non-melanoma skin cancers are rarely fatal when caught early and precancerous lesions are commonly excised before progression, occupational mortality studies and cancer registries are both considered likely to underestimate the true incidence based risk. A substantial supplementary literature examines skin cancer risk among dermatology patients treated therapeutically with topical coal tar (a BaP-containing product) for psoriasis and eczema (e.g., Roelofzen et al., 2010; Stern et al., 1980; Maier et al., 1996), but EPA concluded this literature has limited utility for general population risk assessment, both because of substantial study-design limitations (exposure and outcome misclassification, absent or inappropriate comparison groups, concurrent ultraviolet phototherapy that cannot be disentangled from the coal-tar effect) and because psoriatic skin’s altered anatomy (hyperkeratosis, accelerated desquamation) may reduce PAH penetration relative to normal skin, limiting the generalizability of null or attenuated findings in this population to the general population.

3.4 Reduced Birth Weight and Fetal Growth (Strong Evidence)

Reduced birth weight is supported by convergent findings across four independent birth cohorts using two different exposure-assessment approaches. Using BPDE-DNA adduct biomarkers, Tang et al. (2006), in a Chinese cohort, found that high cord-blood BaP-adduct levels were associated with reduced weight at 18, 24, and 30 months of age, though not at birth itself. In a New York City cohort, Perera et al. (2005b) found no independent association of cord blood BaP adducts or environmental tobacco smoke (ETS) exposure alone with birth weight, but found that a doubling of cord-blood adducts in combination with in utero ETS exposure corresponded to an 8% reduction in birth weight — an interaction finding suggesting that co-exposure to tobacco smoke may potentiate BaP’s developmental effects. Using an independent, dietary-intake-based exposure metric, Duarte-Salles et al. (2012), in a Spanish cohort, found that dietary BaP intake was associated with decreased birth weight, reduced birth length, and small-for gestational-age infants specifically among women with low (below-average) vitamin C intake, and Duarte-Salles et al. (2013), in a larger Norwegian cohort, confirmed a relationship between dietary BaP intake and reduced birth weight and length across the full cohort, including nonsmokers, again with a stronger association among women with lower vitamin C intake — suggesting a possible protective, antioxidant-mediated role for vitamin C and providing biological coherence across independently conducted studies using different populations and exposure assessment methods. Notably, no association with birth length was observed in either the New York City or Chinese adduct-based cohorts, illustrating that even within the “strong evidence” category, not every specific growth parameter is affected consistently across all study populations.

3.5 Neurodevelopmental / Cognitive Deficits (Strong Evidence)

Neurodevelopmental effects are supported by two independent birth-cohort research programs using different exposure-assessment methods but converging on a common conclusion. In the Chinese cohort, Tang et al. (2008) found that increased cord-blood BPDE-DNA adduct levels were associated with decreased developmental quotients in offspring, along with impaired neuromuscular/reflex function. In the New York City cohort, Perera et al. (2009) found that prenatal exposure to a composite measure of eight ambient-air PAHs (a validated proxy incorporating BaP as one constituent) was associated with reduced child IQ at age 5; the same cohort’s longer-term follow-up, Perera et al. (2012b), found associations between prenatal PAH exposure and increased attention problems and anxious/depressed behavior in children assessed at age 6–7. These findings are corroborated by the birth-weight and head-circumference reductions discussed in Section 3.4 (reduced head circumference has itself been linked to neurodevelopmental risk in the broader pediatric literature), providing biological coherence across related developmental endpoints measured in the same cohorts.

3.6 Male Reproductive / Sperm Effects (Suggestive Evidence)

Human evidence of male reproductive effects derives from occupational and clinical populations with elevated PAH-mixture exposure. Coke-oven workers were found to have a substantially higher frequency of oligospermia (19% versus 0% in controls) and twice the rate of morphologically abnormal sperm compared with a non-exposed reference group (Hsu et al., 2006). Elevated BPDE-DNA adduct levels in sperm have been documented in both occupationally PAH-exposed men (Gaspari et al., 2003) and cigarette smokers, and higher sperm-adduct concentrations have been observed specifically in sperm samples not selected for use in intrauterine insemination or in vitro fertilization based on motility and morphology criteria (Perrin et al., 2011a, 2011b) — i.e., adduct burden tracked with independently assessed indicators of poorer sperm quality. Men with higher urinary PAH metabolite concentrations have been found more likely to be infertile in a case-control study (Xia et al., 2009), and an association between BaP-related exposure biomarkers and increased sperm DNA fragmentation (measured by the sperm chromatin structure assay) was reported by Rubes et al. (2010), although EPA’s review noted that the predictive relationship between this particular assay and actual fertility outcomes remains scientifically unsettled. Across this literature, findings are reasonably consistent in direction (worse sperm parameters with higher PAH/BaP biomarker levels) but are limited by an inability to isolate BaP’s specific contribution from the broader PAH mixture and, in several studies, from concurrent cigarette smoking.

3.7 Ischemic Heart Disease (Suggestive Evidence)

Two occupational cohort studies provide the principal human cardiovascular evidence with BaP-specific exposure quantification. In a Canadian aluminum-smelter cohort, Friesen et al. (2010) found elevated IHD risk associated with past cumulative BaP exposure (using a 5-year exposure lag), though the trend did not reach statistical significance, and no association was found with more recent BaP exposure. In a cohort of male asphalt workers, Burstyn et al. (2005) found an elevated, non-statistically-significant association between cumulative BaP exposure and IHD mortality, but a statistically significant trend between average BaP exposure and IHD risk, with approximately a 60% increase in risk between the lowest and highest exposure groups. Both studies relied on statistical exposure-reconstruction models rather than direct individual measurement, and neither could account for co-exposures occurring within the occupational setting (e.g., other combustion byproducts specific to aluminum smelting or asphalt work) or for cardiovascular risk factors and exposures occurring outside the workplace, which is why EPA characterized this evidence as suggestive rather than as supporting a firm conclusion of human cardiovascular hazard.

4. Study Quality Considerations

4.1 Confounding, Particularly by Smoking and Mixture Co-Exposure

The dominant confounding concern throughout the BaP epidemiological literature is that essentially no human population is exposed to BaP in isolation: it is always encountered as one constituent of a complex PAH or combustion-product mixture, and often alongside cigarette smoking, which is itself both a major independent source of BaP exposure and an independent risk factor for many of the same outcomes under study (lung and bladder cancer, cardiovascular disease, reduced fertility, and adverse birth outcomes). The strongest studies in this review explicitly addressed smoking as a confounder in their statistical models (e.g., all three Tier 1 lung cancer cohorts: Armstrong and Gibbs, 2009; Spinelli et al., 2006; Xu et al., 1996), and several developmental studies specifically examined the interaction between BaP-adduct levels and environmental tobacco smoke exposure rather than treating smoking merely as a nuisance covariate (Perera et al., 2005b). Even with statistical adjustment, however, residual confounding by unmeasured or imperfectly measured co-exposures within occupational settings (e.g., other PAHs, metals, or combustion products specific to smelting, coking, or asphalt work) cannot be fully excluded in any single study, which is why EPA’s review places particular weight on the replication of findings across industrially and geographically independent cohorts with different co-exposure profiles (Section 1.4).

4.2 Reverse Causation and Temporality

Reverse causation is a minimal concern for most of the outcomes discussed here because exposure was generally measured before outcome ascertainment: occupational cohort studies track cumulative workplace exposure prospectively or via employment records predating cancer diagnosis or death, and birth-cohort studies measure maternal or cord-blood biomarkers at or before delivery, prior to assessment of birth or developmental outcomes. A partial exception concerns some biomarker-based reproductive studies (e.g., sperm or follicular-fluid BaP levels measured in fertility-clinic patients), where the possibility that an underlying reproductive condition itself alters BaP metabolism, distribution, or clearance cannot be entirely excluded, though there is no specific biological mechanism identified in the literature reviewed here by which infertility would be expected to increase BaP biomarker levels.

4.3 Exposure Misclassification

Exposure misclassification takes several distinct forms across this literature. Occupational cohort studies vary substantially in exposure-assessment rigor, from direct, cumulative BaP air-monitoring data linked to individual work histories (the Tier 1 studies) to cruder job-title or industry-based proxies (many Tier 2 studies); less precise exposure assessment is expected to produce more attenuated and inconsistent effect estimates, which is exactly the pattern EPA’s review identified when comparing Tier 1 and Tier 2 lung and bladder cancer studies. Biomarker studies using BPDE-DNA adducts offer a more direct measure of internal BaP-derived genotoxic dose than air or dietary exposure estimates, but adduct levels reflect recent exposure and metabolic activation capacity (which varies by individual genotype) rather than lifetime cumulative dose, and a single blood or sperm sample may not represent exposure during the etiologically relevant window for outcomes with long latency (e.g., cancer, as opposed to acute developmental exposure windows). Dietary BaP intake estimates (used in the Spanish and Norwegian birth cohorts) depend on food-frequency questionnaire accuracy and on food-composition databases for BaP content, both of which introduce nondifferential measurement error that would generally be expected to bias associations toward the null rather than to create spurious positive findings.

4.4 Co-Exposure to Related Chemicals

Because BaP is both a marker for, and one active constituent of, complex PAH mixtures, essentially every study discussed in this summary reflects some degree of co-exposure to other PAH congeners (and, in occupational and dermal-exposure settings, often to non-PAH combustion products, metals, or other industrial chemicals as well). This is a structural feature of BaP epidemiology rather than a correctable design flaw in any individual study, and it is the central reason EPA’s evidence synthesis repeatedly notes that “the relative contributions of benzo[a]pyrene and of other PAHs cannot be established” even in the strongest available studies. The practical implication is that the epidemiological evidence summarized here should be understood as establishing BaP as a substantial contributing carcinogenic and toxic agent within these mixtures — supported by BaP-specific exposure-response gradients in the strongest studies — rather than as isolating BaP’s independent quantitative contribution to observed health effects apart from its co-occurring mixture partners.

4.5 Study Design and Statistical Power

Several outcome categories in this review (bladder cancer, IHD, melanoma, immune endpoints) are limited by small case counts even in the largest available occupational cohorts, which constrains statistical power to detect a true effect and produces wide confidence intervals that frequently include the null even where the point estimate is elevated — a pattern that should be interpreted as statistical imprecision rather than as evidence of no effect. Conversely, some case-control and cross-sectional developmental and reproductive studies, while adequately powered for their specific outcome, rely on convenience samples (e.g., dermatology clinic patients, fertility clinic patients) that may not be representative of the general population and that can introduce selection bias if enrollment is related to both exposure and outcome status.

5. Weight-of-Evidence Conclusions

EPA’s 2017 IRIS assessment reached the following formal weight-of-evidence conclusions based on its synthesis of the human (and supporting animal) evidence:

• Carcinogenicity: BaP is classified as “carcinogenic to humans” under EPA’s 2005 Guidelines for Carcinogen Risk Assessment, superseding its earlier (1992) classification as a Group B2 “probable human carcinogen.” This conclusion rests on strong human evidence for lung cancer (and, to a lesser degree, bladder and skin cancer) from occupational and environmental PAH-mixture exposure, extensive animal bioassay evidence at multiple tumor sites, an identified and well-characterized mutagenic mode of action, and evidence that the mutagenic key events observed in animals (BPDE-DNA adduct formation, characteristic p53 and K-ras mutation spectra) are also observed in exposed humans, including in lung tumor tissue from PAH-exposed cancer patients.
• Noncancer hazards — developmental toxicity (including developmental neurotoxicity), reproductive toxicity, and immunotoxicity were each identified as human hazards (or potential human hazards) of BaP exposure,

based on a weight-of-evidence integration of the human epidemiological findings summarized in Sections 2–3 together with corroborating, mechanistically coherent animal data.

• Cardiovascular, hepatic, renal, and adult neurological effects were explicitly not concluded to be established human hazards at the time of the 2017 assessment; EPA characterized the evidence for these outcomes as “less robust and consistent” than for the outcomes above and drew no formal hazard conclusion, reflecting the more

limited and more confounded state of both the human and animal evidence for these organ systems (Section 2.6–2.7).

This tiered structure — a small set of outcomes supported strongly enough to drive quantitative dose-response derivation (see RoseyFields’ companion risk assessment methodology page), a broader set of outcomes recognized as probable or established hazards without full quantitative characterization, and a residual set of outcomes for which no conclusion was drawn — mirrors the graded “known/probable/possible/not classifiable” structure used by other expert bodies, including IARC’s Group 1/2A/2B/3 carcinogenicity classifications and EPA’s own A/B1/B2/C/D carcinogen categories referenced throughout the companion ATSDR PAH mixture guidance (Section 3.4 of that document), and functions analogously to “probable link” or “sufficient/limited evidence” determinations issued by other expert panels in comparable toxic-tort and mass-exposure contexts.

6. Use of This Evidence, Alongside Toxicology, in General and Specific Causation

In litigation and regulatory support work, RoseyFields uses the epidemiological evidence summarized above as the primary basis for general causation opinions — whether BaP (or a PAH mixture containing it) is capable of causing the health outcome at issue at exposure levels comparable to those experienced by the relevant population — while pairing it with the mechanistic and animal toxicology evidence (summarized in RoseyFields’ companion risk assessment methodology page) to establish biological plausibility and coherence, consistent with a Bradford Hill type framework (strength and consistency of association, dose-response gradient, temporality, biological plausibility, and coherence across independent lines of evidence).

For lung cancer specifically, the convergence of (a) statistically controlled, BaP-specific exposure-response relationships replicated across three independent occupational cohorts, (b) a well-characterized mutagenic mode of action supported by human biomarker data (BPDE-DNA adducts, characteristic mutation spectra in lung tumors from PAH-exposed individuals), and (c) extensive, concordant animal bioassay evidence, provides an unusually strong general causation foundation. Bladder and skin cancer, and the developmental, reproductive, neurodevelopmental, and immune noncancer outcomes discussed in Sections 2–3, are similarly well supported for general causation purposes, though typically requiring more careful presentation of the underlying exposure response data given smaller case counts, less consistent findings across studies, or (for skin cancer) a more heterogeneous evidence base. For outcomes characterized here as null, inadequate, or unaddressed in the human literature (ovarian toxicity, most immune endpoints, hepatic/renal/adult-neurological effects, general cardiovascular mortality), general causation opinions relying primarily on BaP-specific human epidemiology are not well supported by the current evidence base and would need to rely more heavily on animal and mechanistic data, with appropriately tempered conclusions.

For specific causation, RoseyFields’ analysis connects the general causation evidence above to an individual plaintiff or population through: (1) reconstruction of the individual’s or population’s actual BaP/PAH exposure (source characterization, exposure route and duration, and, where available, biomarker data), compared against the exposure levels and durations reflected in the key studies summarized in Section 3; (2) differential-etiology analysis addressing other known risk factors for the specific outcome at issue, most importantly smoking history given its role as both an independent risk factor and a major BaP exposure source in its own right; (3) consideration of co exposure to other PAH-mixture constituents and, where relevant, application of the ATSDR BaP-equivalent methodology (see RoseyFields’ companion risk assessment methodology page) to characterize the full carcinogenic potency of a measured PAH mixture rather than BaP alone; and (4) attention to life-stage-specific susceptibility, since the strongest and most consistent human evidence for several noncancer outcomes (developmental, neurodevelopmental) specifically concerns prenatal and early-life exposure windows, meaning case-specific timing of exposure relative to pregnancy or early childhood is often a critical factor in whether the epidemiological evidence base actually supports a claimed injury.

References