PFOS Toxicology and Cancer Toxicity

Perfluorooctane Sulfonic Acid (PFOS)

RoseyFields evaluates the toxicology of PFOS using the following evidence base, drawn from the peer-reviewed animal toxicological, mechanistic, and epidemiological literature underlying current PFOS hazard characterization.

1. Chemical Identity and Physicochemical Properties

Perfluorooctane sulfonic acid (PFOS; CASRN 1763-23-1) is an eight-carbon perfluoroalkyl sulfonate (chemical formula C8HF17O3S; molecular weight 500.13 g/mol) historically used as a surfactant, aqueous film-forming foam component, and stain- and oil-repellent treatment. PFOS exists as both linear and branched structural isomers and, at the pH values found in the environment and in the human body, is present almost entirely as the dissociated sulfonate anion rather than the neutral acid. Its physicochemical profile is dominated by the strength of its carbon-fluorine bonds, which render the molecule essentially resistant to biodegradation, hydrolysis, and photolysis.

This resistance to degradation underlies PFOS’s environmental persistence and its capacity for long-range atmospheric and aquatic transport; PFOS has been detected in Arctic biota, including polar bears, far from any point source of contamination (Smithwick et al., 2006), and it is among the most consistently identified perfluoroalkyl substances in global surface waters (Lindstrom et al., 2011). PFOS has low volatility but sorbs appreciably to organic carbon (mean Koc of approximately 1,000 L/kg) (Zareitalabad et al., 2013) and has an estimated log Kow of 4.49, though its amphiphilic (both hydrophobic and oleophobic) character complicates precise octanol-water partitioning measurements. These same properties — persistence, protein affinity, and resistance to metabolic breakdown — are the foundation for the toxicokinetic behavior described below, and in particular for PFOS’s markedly long elimination half-life in humans relative to laboratory animals.

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

2.1 Absorption

Oral bioavailability of PFOS is very high in laboratory rodents; comparative pharmacokinetic studies in rats, mice, and monkeys report serum concentrations after oral gavage that reach or exceed those achieved by intravenous dosing, a pattern consistent with enterohepatic recirculation augmenting apparent absorption (Chang et al., 2012). Because PFOS is only moderately water-soluble and is oleophobic, uptake across cell membranes was long assumed to require membrane transporters; however, more recent biophysical work using planar lipid bilayers found that PFOS’s membrane/water partition coefficients are high enough to support passive diffusion into cells without an obligatory transporter step (Ebert et al., 2020). Cellular uptake also correlates with PFOS’s affinity for phospholipids, which exceeds that of the analogous carboxylic acid PFOA and tracks with accumulation in lung epithelial cell models (Sanchez Garcia et al., 2018). No study has directly quantified absorption in humans, so human uptake is inferred from the ubiquity of PFOS in serum rather than measured directly.

2.2 Distribution

Once absorbed, PFOS distributes throughout the body bound almost entirely to plasma proteins rather than dissolved freely in blood. In vitro binding studies show extensive, high-affinity association with human serum albumin, mediated through van der Waals forces and hydrogen bonding (Chen and Guo, 2009), with additional affinity for low-density lipoproteins (D’Alessandro et al., 2013). Binding strength varies by isomer, with linear PFOS binding calf serum albumin more strongly than several branched isomers (Beesoon and Martin, 2015), and PFOS binding can itself alter albumin’s secondary structure and its capacity to carry endogenous ligands (Liu et al., 2017b). Within cells, PFOS binds liver fatty acid-binding protein, an abundant cytosolic lipid carrier that may facilitate hepatic retention (Zhang et al., 2013b), and hepatic uptake appears to be reinforced by PFOS-induced upregulation of the OATP2 and MRP2 transporters (Yu et al., 2011).

Tissue distribution studies converge on blood and liver as the principal reservoirs of PFOS in both humans and animals (Pérez et al., 2013; Maestri et al., 2006), but there are informative species differences: non-human primates show a nearly 1:1 liver-to-serum accumulation ratio, whereas rodents accumulate substantially more PFOS in liver relative to serum, a distinction relevant to interpreting the higher apparent hepatic burden seen in rodent bioassays. PFOS also crosses the blood-brain and blood-testis barriers in rodents, reaching moderate concentrations in brain and testicular tissue (Qiu et al., 2013; Bogdanska et al., 2011), whereas in adult humans, cerebrospinal fluid levels are roughly two orders of magnitude below serum levels, suggesting more limited penetration of the adult human blood-brain barrier. Plasma concentrations are broadly similar between sexes in short-term rodent studies, though some sex-dependent differences emerge over longer durations and across lifestages.

PFOS also crosses the placenta and is transferred through breast milk, though incompletely relative to maternal serum. Reported ratios of PFOS in cord serum relative to maternal serum generally increase with gestational age (Kato et al., 2014; Porpora et al., 2013), while concentrations in amniotic fluid remain markedly lower than in either placenta or cord blood (Zhang et al., 2013c). Fetal organ concentrations of PFOS have been shown to be lower than in maternal plasma but to rise somewhat across trimesters (Mamsen et al., 2017; Mamsen et al., 2019). Postnatally, PFOS is measurable in breast milk at concentrations roughly one to two orders of magnitude below maternal serum, with breastfeeding duration and maternal serum decline correlated (Cariou et al., 2015; Mondal et al., 2014).

2.3 Metabolism

A consistent finding across the toxicokinetic literature is that PFOS is not metabolized in humans, non-human primates, or rodents. Its strong carbon-fluorine bonds resist enzymatic transformation, so the parent compound persists in the body unchanged until it is excreted. This lack of metabolism is central to interpreting PFOS’s long biological half-life: because clearance cannot proceed through metabolic breakdown, it is governed entirely by excretory pathways, which themselves differ considerably across species.

2.4 Excretion and Half-Life

Urinary excretion is the dominant elimination route in humans, though renal clearance is inefficient: estimated urinary clearance rates are on the order of 0.01-0.04 mL/kg/day and are consistently lower in males than in females (Zhang et al., 2013d; Fu et al., 2016). In rats, by contrast, fecal elimination is comparably important to urinary elimination, and evidence of enterohepatic recirculation — biliary excretion followed by intestinal reabsorption — helps explain why PFOS persists longer than would be predicted from simple renal filtration alone. Early work showed that a bile acid sequestrant increased fecal elimination of radiolabeled PFOS roughly ten-fold in rats (Johnson et al., 1984), and a parallel case report found that bile acid sequestrant treatment in a human subject lowered serum PFOS while increasing stool concentrations (Genuis et al., 2010). In vitro transfection studies confirm that PFOS is a substrate for multiple hepatic and intestinal bile-acid transporters in both humans and rats (Zhao et al., 2015; Zhao et al., 2017), though whether these transporters drive clinically meaningful resorption in vivo remains unresolved.

Half-life estimates vary enormously by species: measured in hours to days in rodents, days to weeks in non-human primates, and years in humans (Fu et al., 2016; Xu et al., 2020c). Human estimates range from roughly one year up to several decades depending on the population and exposure scenario studied, including an estimate of 60.9 years in an occupationally exposed cohort in China (Fu et al., 2016) and shorter estimates, on the order of a few years, in community and newborn populations (Spliethoff et al., 2008; Genuis et al., 2014). Within humans, males consistently show longer half-lives than females, a pattern hypothesized to reflect menstrual blood loss as an additional elimination route unique to females of reproductive age; menstrual serum loss has been estimated to account for a substantial share of the sex difference in elimination (Wong et al., 2014), though subsequent reanalyses have questioned whether this fully explains the effect, raising the possibility of reverse causality or other contributing factors (Ruark et al., 2017; Singer et al., 2018; Lorber et al., 2015). Linear PFOS isomers also appear to have longer half-lives than branched isomers (Xu et al., 2020c). Hair has been identified as an additional, minor elimination route in rats (Gao et al., 2015).

Because of this pronounced interspecies difference in half-life and the absence of metabolic clearance, physiologically based pharmacokinetic (PBPK) models are used to translate administered doses in animal studies into internal (serum or tissue) dose metrics that can be compared across species and extrapolated to humans (Andersen et al., 2008; Thompson et al., 2010a; Gomis et al., 2017). Even so, some tissue-level toxicokinetic behavior — for example, evidence that PFOS in different organs may not follow a single compartmental elimination pattern, with slower clearance from liver than from serum in some studies — indicates that these models simplify a more complex underlying disposition (Iwabuchi et al., 2017).

3. Mechanism(s) of Action

PFOS is not thought to act through a single unifying mechanism; rather, the weight of mechanistic evidence supports multiple, health-outcome-specific modes of action, with the strongest characterization available for hepatic effects and hepatic tumorigenesis.

The most extensively studied pathway involves activation of peroxisome proliferator-activated receptor alpha (PPARα) and related nuclear receptors. In vitro reporter-gene assays show that PFOS can activate PPARα, PPARγ, and, to a lesser extent, PPARβ/δ in both human- and rodent-derived cell lines, though results vary by assay system and PFOS is consistently a weaker PPARα activator than the structurally related PFOA. An adverse outcome pathway has been proposed in which PPARα activation in hepatocytes leads to altered cell-growth signaling (including Kupffer cell activation and downstream cytokine release), increased hepatocyte proliferation with suppression of apoptosis, and eventual clonal expansion of preneoplastic foci culminating in hepatocellular adenomas and carcinomas (Klaunig et al., 2003; Corton et al., 2014). A key uncertainty in applying this pathway to human risk assessment is that constitutive PPARα expression is markedly lower in human liver than in rodent liver, which has historically been used to argue that PPARα-driven rodent liver tumors have limited human relevance for classic PPARα agonists; however, the mechanistic data for PFOS also implicate other, non-PPARα pathways — including constitutive androstane receptor (CAR) activation, suppression of hepatocyte nuclear factor 4-alpha, cytotoxicity, oxidative stress, genotoxicity, and immunosuppression — that are not similarly attenuated in humans, so a purely PPARα-based argument for discounting human relevance is not considered adequate on its own.

For immune effects, the precise mechanism is less well defined, but functional immunotoxicology studies consistently show that PFOS suppresses the antibody response to both T-cell-dependent and T-cell-independent antigen challenges in rodents, implicating effects on B-cell antibody production and, potentially, on T-cell-mediated help for that response. For cardiovascular and lipid effects, disruption of hepatic lipid and lipoprotein metabolism — plausibly linked to the same nuclear receptor pathways implicated in hepatotoxicity — is thought to underlie the serum lipid alterations observed in animal models, including impaired lipoprotein production noted in genetically humanized mouse models of lipid metabolism (Bijland et al., 2011).

4. Animal Toxicological Evidence by Health Outcome

EPA’s underlying systematic review identified the hepatic, immune, cardiovascular, and developmental health outcomes as having the strongest combined weight of evidence (both epidemiological and animal toxicological) for association with PFOS exposure; the animal evidence for these four outcomes is summarized below, along with a brief characterization of the more limited endocrine and neurological animal data.

4.1 Hepatic

The liver is the most consistently identified target organ for PFOS in animal studies, with concordant findings across mice, rats, and monkeys and across short-term, subchronic, chronic, and gestational exposure durations. Increased absolute and relative liver weight has been reported in male BALB/c mice at doses as low as 1.25-5 mg/kg/day for 28 days (Yan et al., 2014) and in male C57BL/6 mice at 2.5-10 mg/kg/day (Xing et al., 2016), with a dose-related increase reaching significance above 0.417 mg/kg/day in a 60-day study (Dong et al., 2011). In Sprague-Dawley rats, significantly increased relative liver weight has been observed at doses as low as 0.15 mg/kg/day in females and 1.33 mg/kg/day in males after 28-day dietary exposure (Curran et al., 2008), with comparable findings in a companion 28-day study (NTP, 2019) and in a subchronic monkey study at 0.75 mg/kg/day (Seacat et al., 2002).

Clinical chemistry changes track the organ-weight findings. Xing et al. (2016) reported dose-dependent increases in serum ALT of 50% and 88% above control at 5 and 10 mg/kg/day in male mice, together with a biologically significant increase in gamma-glutamyltransferase (up to 535% at the highest dose), while Yan et al. (2014) observed even larger ALT elevations (over 700% at 5 mg/kg/day) in a similarly designed study, without a clear linear dose-response across the narrower dose range tested. Histopathological confirmation is available from several studies: a 28-day dietary study in Sprague-Dawley rats found dose-dependent hepatocellular hypertrophy and cytoplasmic vacuolation/alteration in both sexes, reaching 10/10 animals affected at the highest dose tested (NTP, 2019), and a companion 28-day study identified zone 3 hepatocellular hypertrophy in males at 3.21 and 6.24 mg/kg/day (Curran et al., 2008). In the only chronic bioassay available, hepatocellular hypertrophy was significantly increased in both sexes across the full dose range tested (0.098-0.984 mg/kg/day in males; 0.299-1.251 mg/kg/day in females) over 103 weeks, with a clear positive dose-response relationship (Butenhoff et al., 2012).

4.2 Immune

PFOS has been repeatedly shown to alter immune organ weights and, more importantly, to functionally suppress the antibody response in rodents. Reductions in thymus and/or spleen weight have been reported across studies of 7 to 60 days’ duration in both mice and rats, generally most pronounced at higher doses and sometimes exhibiting sex-specific sensitivity — for example, female Sprague-Dawley rats showed reduced absolute thymus weight at doses as low as 1.25 mg/kg/day in a 28-day study, a lower threshold than observed in males from the same study (NTP, 2019).

The most consequential immune finding is suppression of the antibody response to sheep red blood cells (SRBC), a T-cell-dependent immune challenge measured via the plaque-forming cell (PFC) IgM response. Peden-Adams et al. (2008) found this response suppressed in both male and female B6C3F1 mice following 28-day oral exposure, with effects detectable at doses as low as 0.0017 mg/kg/day in males (a lower threshold than in females, at 0.017 mg/kg/day) — among the most sensitive endpoints identified for PFOS in the animal database. The same study found reduced IgM titers following a T-cell-independent antigen challenge in females, indicating that PFOS can act on antibody-producing B cells directly, not solely through T-cell help. Dong et al. (2009, 2011) similarly reported dose-dependent reductions in the SRBC-specific IgM PFC response and in serum IgM in male C57BL/6 mice following 60-day exposure, though delayed-type hypersensitivity responses (footpad swelling) were not consistently altered in the same animals, and a parallel rat study using a different T-dependent antigen found no significant change in antibody titers or footpad swelling (Lefebvre et al., 2008), underscoring some inconsistency across assay types and species.

4.3 Cardiovascular

Direct cardiac effects of PFOS in animal studies have been comparatively minimal. Heart weight changes have been reported in some studies but are frequently confounded by concurrent decreases in body weight (Xia et al., 2011; NTP, 2019), and short-duration studies are not well suited to detecting slower processes such as atherosclerotic plaque formation. Gestational exposure studies provide somewhat stronger signals: PFOS exposure in drinking water during gestation increased ventricular wall thickness and cardiac fibrosis modestly in exposed dams’ offspring and increased maternal blood pressure late in gestation (Dangudubiyyam et al., 2022), and gestational gavage exposure in rats produced elevated systolic blood pressure in adult offspring of both sexes, an effect associated in males with a reduced nephron number at postnatal day 22 (Rogers et al., 2014).

The most consistent cardiovascular-relevant finding in animals is altered serum lipid homeostasis: decreased total cholesterol, triglycerides, and/or high-density lipoprotein have been reported in non-human primates (Seacat et al., 2002), rats (Curran et al., 2008; NTP, 2019), and mice (Yan et al., 2014; Bijland et al., 2011) across a range of PFOS doses. This finding is notable chiefly for the direction in which it diverges from the human epidemiological literature, discussed further in Section 6 below.

4.4 Developmental and Reproductive

Developmental toxicity is well documented in mice, rats, and rabbits following gestational PFOS exposure. Maternal effects — decreased body weight or weight gain during gestation — have been observed at higher doses (approximately 2-20 mg/kg/day) in CD-1 mice (Thibodeaux et al., 2003; Lee et al., 2015) and Sprague-Dawley rats (Thibodeaux et al., 2003), and in New Zealand white rabbits at 1 mg/kg/day and above (Argus Research Laboratories, 2000), while several lower-dose studies (generally at or below approximately 3 mg/kg/day) found no significant maternal body weight effects (Lai et al., 2017a; Wan et al., 2020; Fuentes et al., 2006; Butenhoff et al., 2009; Zhang et al., 2021; Conley et al., 2022).

Fetal and pup viability effects appear most pronounced in mice: Lee et al. (2015) reported significantly increased resorptions, post-implantation loss, and dead fetuses at 2 and 8 mg/kg/day in CD-1 mice, and Thibodeaux et al. (2003) found increased post-implantation loss at 20 mg/kg/day in the same strain, whereas parallel rat studies at comparable or higher doses generally did not show effects on implantation or live fetal counts (Thibodeaux et al., 2003; Zhang et al., 2021; Butenhoff et al., 2009; Conley et al., 2022), suggesting mice may be more sensitive than rats for this endpoint. In rabbits, PFOS produced a dose-related increase in abortion incidence, reaching 42.8% of does at 3.75 mg/kg/day (Argus Research Laboratories, 2000). Reduced postnatal pup viability has also been documented in one- and two-generation reproductive toxicity studies in Sprague-Dawley rats, with a significant decrease in the pup viability index (survival from birth to postnatal day 5) at maternal doses of 1.6 mg/kg/day and complete litter loss in some dams at 2-3.2 mg/kg/day (Luebker et al., 2005a; Luebker et al., 2005b). Developmental studies have also identified windows of heightened sensitivity, for instance gestational days 19-21 as a critical period for neonatal lung maturation and survival in rats.

4.5 Endocrine and Neurological (Supporting Evidence)

Endocrine and nervous system outcomes were characterized in the underlying assessment as having evidence “suggestive of” but not fully meeting criteria for an established association with PFOS exposure, based on a combined body of high- and medium-confidence animal and epidemiological studies. On the endocrine side, animal data point to disruption of normal thyroid function, including decreased circulating free and total thyroxine (T4) and total triiodothyronine (T3), alongside less consistent histopathological changes in the thyroid gland and reductions in some hypothalamic-pituitary-adrenal axis hormones. For the nervous system, animal studies provide moderate but inconsistent evidence of effects on learning and memory and on neurotransmitter concentrations. Because these two outcomes were not prioritized for full quantitative dose-response analysis in the underlying assessment, individual animal study citations for these endpoints were not extracted for this summary; RoseyFields notes them here as areas where the hazard signal exists but the animal database is comparatively less mature than for the four outcomes above.

Renal and reproductive (non-developmental) outcomes were similarly not prioritized for detailed synthesis in the core assessment reviewed here, and the animal toxicological database for these outcomes was correspondingly less developed at the time of this summary’s preparation.

5. Animal Carcinogenicity Bioassay Findings

The animal carcinogenicity database for PFOS rests on a single chronic bioassay: male and female Crl:CD(SD) rats were fed diets containing 0, 0.5, 2, 5, or 20 ppm PFOS for 103-104 weeks (Thomford, 2002; published in part as Butenhoff et al., 2012). This study identified statistically significant, dose-related increases in hepatocellular adenomas in high-dose males (7/43, 16%, versus 0/41 in controls) and in high-dose females (5/31, 16%, versus 0/28 in controls), with combined hepatocellular adenomas and carcinomas reaching 19% (6/32) in high-dose females compared with 0% in controls; positive trend tests were statistically significant (p < 0.01) in both sexes. A single hepatocellular carcinoma was identified in a high-dose female, a tumor type described as rare for this rat strain.

In addition to liver tumors, the same bioassay found a statistically significant positive trend (Cochran-Armitage test, p < 0.05) for pancreatic islet cell carcinomas in male rats, with combined islet cell adenomas and carcinomas rising from 11% in controls to 20% at the highest dose, although pairwise comparisons with control at individual doses did not reach statistical significance. Thyroid C-cell adenomas and mammary gland tumors (chiefly fibroadenoma/adenoma) were also observed but did not show a dose-response relationship and, for thyroid tumors, occurred at their highest incidence in the control group, arguing against a treatment-related effect for these two tumor types.

Based on this bioassay together with mechanistic and epidemiological evidence, PFOS has been classified as “Likely to Be Carcinogenic to Humans” under the framework of EPA’s 2005 Guidelines for Carcinogen Risk Assessment. This classification reflects: (1) a positive tumor response across more than one sex and tumor site within a single well-conducted chronic study; (2) a rare tumor type (hepatocellular carcinoma) observed in a single experiment, judged relevant to humans; and (3) support from mechanistic data consistent with tumor-promoting processes — nuclear receptor activation, cytotoxicity, genotoxicity, oxidative stress, and immunosuppression — that plausibly relate to the observed tumor response. This conclusion is bolstered by site concordance with more recent human epidemiological studies reporting increased liver cancer risk associated with PFOS exposure (Cao et al., 2022; Goodrich et al., 2022), discussed further below.

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

Across health outcomes, quantitative toxicity values (points of departure and reference doses) derived from epidemiological studies are consistently lower — by roughly two to three orders of magnitude — than those derived from animal toxicological studies of the same health outcome. In practical terms, adverse associations in human populations tend to emerge at serum PFOS concentrations corresponding to much lower administered-dose equivalents than the doses required to produce comparable effects in laboratory rodents. Several factors plausibly contribute to this divergence, each with direct relevance to how RoseyFields should weigh the animal database against the human evidence.

First, and most fundamentally, toxicokinetic differences between species are substantial: PFOS’s elimination half-life is measured in hours to days in rodents but in years in humans (see Section 2.4), meaning that a given administered dose in an animal study corresponds to a much lower cumulative internal (serum) dose than the same nominal dose would produce in a human, whose slower clearance allows PFOS to accumulate to much higher steady-state serum concentrations from comparatively low daily intakes. PBPK modeling is used to bridge this gap by converting animal administered doses and human epidemiological serum concentrations into a common internal-dose metric, but these models carry their own uncertainty and may not fully capture species-specific differences in excretion, particularly the fecal/enterohepatic pathways that appear more prominent in rodents than in humans.

Second, animal toxicological studies are necessarily conducted at administered doses far above typical background human environmental exposure levels, introducing uncertainty when extrapolating dose-response relationships down to the much lower exposure range relevant to the general population. This is illustrated concretely by the serum lipid findings summarized in Section 4.3: animal studies conducted at comparatively high doses generally show decreased serum cholesterol and triglycerides, sometimes following an apparent biphasic pattern in which lower doses increase lipid levels and higher doses decrease them (Yan et al., 2014), whereas human epidemiological studies conducted at much lower, environmentally relevant PFOS exposures have more often reported positive associations between PFOS and total cholesterol (Eriksen et al., 2013). This directional discordance for a well-studied endpoint is a clear example of how dose range, and not just species, can drive different observed relationships, and it is compounded by methodological factors specific to rodent studies, including inconsistent fasting protocols and the lower fat content of standard laboratory rodent chow relative to typical human diets.

Third, there may be genuine mechanistic differences between species in how PFOS produces its effects. The clearest example is PPARα-mediated hepatotoxicity and hepatocarcinogenesis: rodent liver expresses substantially higher levels of PPARα than human liver, and the extent to which PPARα activation contributes to the human hepatic response to PFOS (as opposed to rodent-specific pathways) is not fully resolved, even though, as discussed in Section 3, multiple non-PPARα mechanisms are also implicated and are not similarly attenuated in humans. Similar mechanistic uncertainty likely applies, to varying degrees, across other health outcomes.

For cancer specifically, the degree of concordance between the animal bioassay and human epidemiology varies by tumor site. Liver cancer shows meaningful concordance: the chronic rat bioassay identified hepatocellular tumors, and more recent human studies have reported statistically significant associations between PFOS exposure and hepatocellular carcinoma, including a nested case-control study within the Multiethnic Cohort (Goodrich et al., 2022) and a Chinese case-control study (Cao et al., 2022), lending mutual support to the two lines of evidence. By contrast, the pancreatic islet cell tumors identified in the rat bioassay do not have a clearly corresponding human epidemiological signal, while some human studies report associations between PFOS and kidney cancer (Shearer et al., 2021) or, in a Swedish cohort with high drinking-water exposure, mixed findings across several cancer sites (Li et al., 2022) without a clear corresponding tumor type in the single available animal bioassay. EPA’s own guidance recognizes that site concordance between animal and human carcinogenicity findings is not required for an agent to be judged carcinogenic in both species, since tumor sites can differ between species for a variety of toxicokinetic and mechanistic reasons; nonetheless, this lack of full concordance is a source of residual uncertainty in extrapolating specific tumor-site risk from animals to humans.

Given these combined uncertainties in extrapolating from experimental animal doses and mechanisms to human risk, EPA’s own toxicity value derivation approach — which RoseyFields’ evidence base draws on here — generally prioritized medium- and high-confidence epidemiological studies over animal toxicological studies when both were available for a given health outcome, treating the animal database as an important source of hazard identification, mechanistic support, and corroboration (particularly for cancer), while relying on human data where available to anchor quantitative dose-response conclusions.

7. Data Gaps and Areas of Ongoing Research

Several important uncertainties remain in the PFOS animal and human toxicological database that bear on the confidence RoseyFields places in specific conclusions above:

Endocrine and nervous system effects. Both outcomes currently sit in an intermediate evidentiary category — more than merely suggestive, but not yet meeting the threshold applied to the four priority outcomes (hepatic, immune, cardiovascular, developmental). Thyroid hormone disruption in animals and humans, and neurodevelopmental signals in children (including some evidence for altered cognitive and behavioral outcomes), warrant continued monitoring as the literature matures.

Renal and reproductive (non-developmental) outcomes. These health outcomes have a comparatively less developed animal toxicological literature relative to the four priority outcomes discussed in Section 4, despite some epidemiological signals (for example, kidney cancer associations in certain cohorts, Shearer et al., 2021) that merit continued animal mechanistic and dose-response investigation.

Mode-of-action uncertainty for carcinogenicity. While multiple plausible mechanisms (PPARα and other nuclear receptor activation, cytotoxicity, genotoxicity, oxidative stress, immunosuppression) are consistent with the observed hepatic and pancreatic tumor findings, no single mode of action has been fully established, and the relative contribution of each pathway — and its quantitative relevance to human risk — remains an active area of investigation.

Isomer-specific toxicity and toxicokinetics. Linear and branched PFOS isomers differ in protein-binding affinity, half-life, and placental transfer efficiency, but most toxicological studies (particularly older ones) did not characterize isomer composition, limiting the ability to attribute effects to specific isomeric forms or to refine human relevance judgments accordingly.

Serum lipid dose-response discordance. The apparent directional difference between animal studies (generally showing decreased serum lipids) and human epidemiology (more often showing positive associations with cholesterol) at different points along the exposure range is not fully resolved and may reflect a combination of dose-range effects, species differences in lipid handling, and rodent-study methodological factors (fasting status, diet composition).

Susceptible populations and exposure windows. Critical windows of developmental vulnerability (for example, specific gestational-day windows affecting neonatal lung maturation) have been identified for some endpoints but not systematically characterized across the full range of health outcomes, and the animal and human literature has not yet adequately characterized whether individuals with pre-existing conditions, differing genetic backgrounds, or disproportionate co-exposure to other environmental stressors experience heightened sensitivity to PFOS.

Toxicokinetic model refinement. Existing PBPK models used to bridge animal and human dosimetry rely on simplifying assumptions about compartmental behavior and species-specific transporter function; as noted in Section 2, some tissue-level data are not well described by standard compartmental models, indicating room for refinement as mechanistic transporter data accumulate.

References