The ATSDR Toxicological Profile for Perfluoroalkyls (2021), the IARC monograph programme, and the C8 Science Panel converge on a small, specific set of human health endpoints linked to PFOA and PFOS exposure. This brief walks through what those bodies actually concluded, in their own language, without scaremongering or extrapolation.
As of the 2023 IARC re-evaluation (Monographs Vol 135), PFOA is classified Group 1, carcinogenic to humans, and PFOS is Group 2B, possibly carcinogenic to humans. Independently, the C8 Science Panel (2011–2013) found "probable links" between PFOA exposure and six specific conditions: kidney cancer, testicular cancer, ulcerative colitis, thyroid disease, pregnancy-induced hypertension, and hypercholesterolaemia.
Everything below unpacks those two sentences against the underlying ATSDR, IARC, EFSA, and C8 documents, and tells you which endpoints have strong epidemiological backing and which are still contested. If you want to know where these molecules come from before reading about their effects in people, the EPA April 2024 PFAS MCL guide covers the regulatory side, and the UCMR 5 explainer covers the occurrence data.
Human exposure to long-chain perfluoroalkyls is dominated by three routes, ranked by contribution in most non-occupationally-exposed populations: drinking water, diet, and indoor dust. ATSDR's 2021 Toxicological Profile for Perfluoroalkyls (the agency is the Agency for Toxic Substances and Disease Registry, an arm of the CDC) summarises the evidence base across roughly 800 pages, and the dominant route in any individual depends almost entirely on whether their tap water is contaminated.
For people drinking PFAS-impacted water, drinking water typically accounts for the majority of total intake, often 70–90%, because PFOA and PFOS do not break down in conventional treatment and pass straight through chlorination, sand filtration, and disinfection. Diet contributes through bioaccumulation in fish, dairy, eggs, and (historically) microwave popcorn bags and grease-resistant food packaging. House dust contributes through abrasion of treated carpets, upholstery, and textile finishes, measurable but minor for adults, and a larger fraction for crawling infants and toddlers who ingest more dust per kilogram of body weight.
Once inside the body, long-chain PFAS bind to serum albumin and to liver fatty-acid binding proteins, distribute primarily to blood serum and liver, and are excreted slowly through the kidneys. ATSDR reports geometric mean serum half-lives of roughly 2.7 years for PFOS, 3.5 years for PFOA, and 8.5 years for PFHxS in adults, orders of magnitude longer than almost any other industrial chemical with comparable molecular weight. The practical consequence is that body burden integrates exposure over years to decades, and that removing the source (filtering tap water with an NSF 53 P473 or NSF 58 certified system) reduces serum levels on a half-life rather than a flush timescale.
Placental and lactational transfer have been measured directly: cord blood routinely shows PFOA and PFOS concentrations roughly 30–80% of maternal serum, and breast milk transfers at roughly 1–2% of maternal serum, small per feed, but compounded over months. The implication, which EFSA and ATSDR both treat as central, is that the fetus and breast-fed infant are exposed throughout the windows the developmental endpoints below concern.
The C8 Science Panel is the single largest population study of PFOA health effects ever conducted, and it remains the foundation of almost every regulatory conclusion about PFOA in humans. It was created in 2005 as part of the settlement of a class-action lawsuit (Leach v. E.I. du Pont de Nemours) against DuPont's Washington Works plant in Parkersburg, West Virginia, whose PFOA emissions had contaminated public and private water supplies along the Ohio River. The settlement funded the panel, gave it access to the affected population (~69,000 people who provided blood samples and health questionnaires through the C8 Health Project), and committed DuPont to medical monitoring for whichever conditions the panel found "probably linked" to PFOA exposure.
Three epidemiologists, Tony Fletcher, Kyle Steenland, and David Savitz, were independent of both the plaintiffs and the defendant, and worked under a binding protocol that defined "probable link" as more likely than not that PFOA exposure was capable of causing the condition at the doses observed in the affected community. Between 2011 and 2013 they released a series of reports covering dozens of candidate endpoints. Six cleared the probable-link threshold; the rest did not.
| Condition | Endpoint type | Direction of association |
|---|---|---|
| Kidney cancer | Cancer | Increased risk with cumulative PFOA serum exposure |
| Testicular cancer | Cancer | Increased risk with cumulative PFOA serum exposure |
| Ulcerative colitis | Autoimmune / GI | Increased incidence at higher serum quartiles |
| Thyroid disease | Endocrine | Increased incidence, both hypo- and hyperthyroid |
| Pregnancy-induced hypertension | Reproductive / cardiovascular | Increased incidence including pre-eclampsia |
| Hypercholesterolaemia | Lipid metabolism | Higher total and LDL cholesterol with higher serum PFOA |
Two things about that table are worth restating because they are routinely overstated in secondary coverage. First, "probable link" is a defined legal-scientific term meaning more-likely-than-not, not "proved." Second, the panel did not find a probable link for several conditions where popular reporting often assumes they did, including breast cancer, prostate cancer, type 2 diabetes, and most birth-defect categories. The full list of negative findings is in the panel's 2012 status report, and the discipline of distinguishing the six positives from the much longer list of negatives is what gives the C8 work its weight.
For the longer history of how the C8 work came about, including the Parkersburg cattle die-off and the internal DuPont documents that surfaced through discovery, the C8 / DuPont guide walks through the litigation timeline.
The International Agency for Research on Cancer (IARC), part of the World Health Organization, evaluates carcinogenic hazard in its Monographs series. IARC categories rank the strength of evidence that an agent can cause cancer, not the magnitude of risk at any given exposure. The categories are: Group 1 (carcinogenic to humans), Group 2A (probably carcinogenic), Group 2B (possibly carcinogenic), Group 3 (not classifiable), and Group 4 (probably not carcinogenic, used only once, for caprolactam).
IARC has evaluated PFOA twice. Monograph Vol 110, published in 2017 (with the working-group meeting in 2016), placed PFOA in Group 2B, possibly carcinogenic to humans, based on limited evidence in humans for cancer of the testis and kidney and limited evidence in experimental animals. PFOS was not evaluated in Vol 110.
The 2023 re-evaluation, Monograph Vol 135, took both compounds in a single working group. The conclusions, published as an IARC press release on 30 November 2023 and detailed in the monograph itself, were:
It is worth being precise about what Group 1 means and does not mean. IARC Group 1 is a hazard classification: it asserts that PFOA can cause cancer in humans under some circumstances of exposure. It is not a statement that any particular drinking-water exposure level will cause cancer in any particular person, and it does not specify a dose–response or a population attributable fraction. The same Group 1 list includes tobacco smoking, ionising radiation, processed meat, and outdoor air pollution, agents with vastly different real-world risk magnitudes.
The European Food Safety Authority (EFSA) published its Scientific Opinion on the risk to human health related to the presence of perfluoroalkyl substances in food in September 2020 (EFSA Journal 2020;18(9):6223). EFSA's panel reviewed PFOA, PFOS, PFNA, and PFHxS together and concluded that the most sensitive human endpoint, the one occurring at the lowest internal exposures, was reduced antibody response to routine childhood vaccines.
The evidence base for this conclusion comes principally from a series of birth cohort studies in the Faroe Islands led by Philippe Grandjean, plus replication in cohorts in Norway, Germany, and the United States. The Faroese studies measured antibody titres to diphtheria and tetanus vaccines at ages 5 and 7, and found that children with higher prenatal and early-childhood PFAS serum concentrations produced statistically lower antibody responses to the same vaccine schedule. The effect was dose-dependent and observed across all four PFAS that EFSA evaluated.
EFSA used this endpoint to derive a tolerable weekly intake (TWI) of 4.4 ng per kg of body weight per week for the sum of PFOA, PFOS, PFNA, and PFHxS, orders of magnitude lower than its previous (2018) provisional TWIs for the individual compounds. The TWI is calibrated to keep serum PFAS levels in 1-year-olds below the concentration at which a 10% reduction in measles vaccine antibody response was observed in the cohort data. We unpack what the TWI means for pregnancy and infant exposure in the tolerable intake guide.
The immune-suppression finding is not limited to vaccine response. ATSDR's 2021 profile catalogues additional associations with reduced lymphocyte counts, altered cytokine profiles, and increased self-reported infectious disease incidence in PFAS-exposed cohorts. The C8 panel's probable link to ulcerative colitis sits in the same immune-modulation cluster, since UC is a disease of dysregulated mucosal immunity.
The developmental endpoints best supported by the human epidemiology are reduced fetal growth (measured as lower birth weight at a given gestational age) and delayed onset of menarche in girls with higher prenatal PFOA exposure. Both endpoints appear consistently across multiple cohorts, with effect sizes that are statistically robust but modest in absolute terms.
For birth weight, ATSDR's 2021 profile pools results from more than thirty cohort studies and reports a dose-dependent inverse association: each 1 ng/mL increase in maternal serum PFOA is associated with roughly a 10–20 gram reduction in birth weight, controlling for gestational age, maternal weight gain, smoking, and other standard covariates. The mechanism is not fully resolved, proposed pathways include altered placental thyroid hormone handling, vascular effects on placental perfusion, and lipid-metabolism interactions, but the association itself replicates across populations and study designs. EPA's June 2022 interim updated Health Advisory for PFOA (cited at 0.004 ng/L for the lifetime HA) was driven primarily by the developmental and immune endpoints rather than by cancer.
For menarche, the C8 Health Project data and several European cohorts find that girls in the highest PFOA exposure quartiles reach menarche on average 4–6 months later than girls in the lowest quartiles. The clinical significance of a 4–6 month delay is modest at the individual level; the public-health significance is in what it implies about endocrine disruption during a critical developmental window.
Outcomes that are not consistently supported by the human data, despite occasional positive single studies, include most congenital malformations, autism spectrum disorder, ADHD, and gestational diabetes. The honest summary is: the developmental signal is real and reproducible for growth and pubertal timing, and weaker or absent for neurodevelopmental and metabolic outcomes.
In June 2022, the US EPA released updated interim drinking-water Health Advisories (HAs) for PFOA at 0.004 ng/L and PFOS at 0.02 ng/L. Both numbers are well below current analytical detection limits, which means they are not enforceable concentrations, they are a regulatory signal that EPA's risk assessors believed the safe lifetime intake was lower than instrumentation could measure.
The interim HAs were superseded for compliance purposes by the April 2024 National Primary Drinking Water Regulation under 40 CFR §141.61(c), which sets enforceable MCLs at 4.0 ng/L for both PFOA and PFOS (with maximum contaminant level goals, MCLGs, of zero, reflecting the no-threshold treatment EPA gives to known and likely human carcinogens). Utilities must complete initial monitoring by 2027 and meet the MCLs by 2029. The 4.0 ng/L MCL is the lowest analytical concentration that EPA considers feasible to measure reliably under Method 533 / Method 537.1 in a regulatory compliance context, it is a feasibility floor, not a safety threshold.
The epidemiology has real limits that responsible reporting should name explicitly. First, almost every cohort study is observational, and unmeasured confounders (socioeconomic status, co-exposure to other chemicals, dietary patterns) cannot be ruled out by statistical adjustment alone. Second, the C8 cohort was geographically clustered and disproportionately Appalachian, generalising to populations with different baseline disease rates and exposure routes requires caution. Third, most PFAS health research focuses on PFOA and PFOS specifically; the toxicology of replacement compounds like GenX (HFPO-DA), PFBS, and the broader thousands-of-PFAS universe is far thinner. Fourth, dose–response curves at the low-ng/L drinking water exposures most of the US population experiences are extrapolations downward from cohorts with serum levels often 5–50× the current population median.
None of these caveats undermines the Group 1 classification or the six probable links. They do mean that quantitative individual-risk estimates, "your tap water gives you an X% chance of kidney cancer", are not what the science currently supports. The science supports population-level statements of association and hazard, and it supports the regulatory conclusion that long-chain PFAS exposure should be reduced. That is the frame EPA, ATSDR, IARC, EFSA, and the C8 panel all converge on.
If you want to know what's in your own tap, the UCMR 5 explainer covers the federal occurrence dataset and the home PFAS testing guide covers the certified labs that will run EPA Method 537.1 / 533 on a tap sample.
One ZIP, one PDF brief. Pregnancy/infant callout if applicable.
Pull my report, $15