The following is a sample from a larger document.
This section is for: Literature Audit
The section is from a Focused Issue Brief on: Assessing Scientific Causation Claims in PFAS Exposure Litigation
The primary research jurisdiction is: United States of America
[Back]
Literature Audit
The peer-reviewed literature on PFAS exposure and human health outcomes spans thousands of studies published over three decades, but this volume creates a misleading impression of scientific consensus. Most published research focuses on exposure measurement, environmental fate, or animal toxicology rather than establishing causation for specific human diseases. When studies do examine human health effects, they vary enormously in methodological quality, study design, and the strength of conclusions their data can actually support.
This audit examines what the literature establishes about causal relationships between PFAS exposure and human disease, assessed against epidemiological standards for demonstrating causation. The focus is narrow: studies that attempt to link measured or estimated PFAS exposure in humans to specific adverse health outcomes, particularly the conditions most commonly alleged in current litigation: kidney cancer, testicular cancer, ulcerative colitis, thyroid disease, pregnancy-induced hypertension, and elevated cholesterol.
Study Design and Methodological Quality
Some of the stronger evidence comes from large prospective cohort studies that follow exposed populations over time and compare disease rates across exposure gradients or to lower-exposure comparison groups. The most frequently cited research includes the C8 Health Studies, which followed approximately 69,000 residents in the Ohio River Valley with documented PFOA exposure from DuPont’s Washington Works facility. These studies benefit from measured exposure data rather than estimates, substantial sample sizes, and follow-up periods extending over decades.
Even these well-designed studies face methodological constraints that limit causal inferences. The C8 population experienced exposure to multiple PFAS compounds simultaneously, making it impossible to attribute health effects to any single chemical. Exposure levels varied dramatically within the study population. Some participants had serum PFOA concentrations 100 times higher than background levels, while others showed only modest elevation above typical U.S. population levels. The studies cannot determine whether observed associations reflect effects from peak exposures, cumulative exposure over time, or exposures during specific vulnerable periods.
Cross-sectional studies, which measure both exposure and disease status at a single point in time, comprise the majority of published PFAS health research. These studies can identify associations but cannot establish temporal sequence—whether exposure preceded disease or disease influenced exposure patterns. Single serum measurements characterize exposure in most cross-sectional studies, and serum levels can be influenced by recent exposures, elimination rates, and physiological factors that may not reflect the historical exposures most relevant to disease development.
Case-control studies compare PFAS levels between people who have developed disease and matched controls who have not. Many PFAS case-control studies face important methodological limitations. Serum samples collected after disease diagnosis form the basis for many studies, but cancer treatment, kidney disease, and other conditions can alter PFAS elimination rates and blood concentrations.
Studies that find higher PFAS levels in cancer patients than controls may detect the effect of disease on PFAS metabolism rather than the effect of PFAS exposure on disease risk.
Animal toxicology studies consistently show adverse effects from PFAS exposure at high doses, but these findings face standard challenges in extrapolating from laboratory animals to humans. Rodent studies often use exposure doses substantially higher than typical human exposures. Rodents eliminate PFAS compounds much more rapidly than humans. PFOA has a half-life of hours to days in rodents compared to 3-4 years in humans. Equivalent serum concentrations therefore represent vastly different exposure patterns between species.
Sample Size and Statistical Power
Some systematic reviews have noted that the evidence base for PFAS exposure and kidney cancer is limited, with relatively few studies meeting stringent criteria for exposure assessment and statistical power, and with mixed findings across studies.
Studies that report positive associations often rely on subgroup analyses that substantially reduce sample sizes. A frequently cited study linking PFAS exposure to testicular cancer found an association only when researchers restricted the analysis to men diagnosed before age 40, reducing the effective sample size from 720 to 87 cases. Post-hoc analysis of this type dramatically increases the likelihood of false positive findings, particularly when researchers examine multiple endpoints and subgroups simultaneously.
Investigators routinely test associations between multiple PFAS compounds and multiple health outcomes in the same dataset, then report statistically significant findings without adjusting for the number of comparisons performed. A study that examines five PFAS compounds and ten health outcomes conducts fifty statistical tests. Even with no true associations, chance alone would produce 2-3 statistically significant findings at the conventional p<0.05 threshold.
Exposure Assessment Quality
The reliability of exposure assessment varies dramatically across studies and fundamentally determines what conclusions researchers can draw. The most rigorous studies use measured serum concentrations from samples collected prior to disease diagnosis, but such prospective biomarker studies remain relatively rare due to their cost and logistical complexity.
Estimated exposures based on residential proximity to PFAS sources, occupational histories, or consumption of contaminated food and water form the basis for most studies. These exposure estimates introduce substantial measurement error that typically biases results toward finding no association, even when true effects exist. Some estimation methods systematically overestimate exposure in certain populations, potentially creating spurious positive associations.
Studies of occupational exposure face particular methodological challenges. Workers in fluorochemical manufacturing facilities experience complex, time-varying exposures to multiple PFAS compounds through inhalation, dermal contact, and incidental ingestion. Job titles or department assignments estimate exposure in most occupational studies rather than personal monitoring data. This approach misses substantial variation in actual exposure within job categories and cannot account for changes in manufacturing processes, safety protocols, or chemical formulations over time.
Community exposure studies around contaminated sites often use distance from the pollution source as a proxy for exposure level. This approach ignores groundwater flow patterns, historical changes in contamination sources, residential mobility, and individual differences in water consumption, diet, and lifestyle factors that affect PFAS uptake and elimination.
Funding Sources and Research Design
Corporate funding has shaped PFAS research in ways that are visible in study design, endpoint selection, and interpretation of findings. During the early PFAS health literature, industry funding (including by major manufacturers) supported a meaningful share of published studies through direct grants and contract research organizations.
Critics argue that industry-funded studies are more likely to report null findings than independently funded research, though the extent of this pattern depends on study selection and endpoints.
This pattern reflects several mechanisms. Industry-funded studies tend to use study designs that are less likely to detect health effects: shorter follow-up periods, broader exposure categories that dilute dose-response relationships, and endpoint definitions that exclude borderline or subclinical disease. Industry studies are also more likely to adjust for covariates that may actually be intermediate steps in the causal pathway from exposure to disease, thereby removing rather than controlling for confounding.
The 3M-funded study of perfluorooctane sulfonate (PFOS) exposure in company employees illustrates these patterns. The study followed workers for an average of only six years after hire, insufficient time for most cancers to develop from occupational exposures. The analysis grouped all production workers together regardless of their specific job duties or measured exposure levels, obscuring dose-response relationships. Some secondary analyses and critiques have argued that alternative modeling choices, longer follow-up, or different exposure categorizations could change inferences drawn from certain occupational datasets.
Government funding has also influenced research directions through different mechanisms. EPA-funded research has focused heavily on environmental fate and transport rather than health effects, reflecting the agency’s regulatory priorities and technical expertise. NIH funding for PFAS health research increased dramatically after 2018 but has concentrated on mechanistic studies and biomarker development rather than large-scale epidemiological investigations.
The military’s substantial investment in PFAS research stems from contamination at hundreds of defense installations where firefighting foams were used for decades. Department of Defense-funded studies tend to focus on acute exposure scenarios and occupational health effects rather than chronic low-level community exposures.
Citation Patterns in Litigation and Regulatory Proceedings
Expert reports and regulatory filings in PFAS litigation often reflect selective citation patterns that can overstate what the literature establishes. Plaintiff-side experts routinely cite preliminary studies, conference abstracts, and research that examined multiple endpoints without noting negative findings for other health outcomes in the same studies.
Some studies examining PFAS exposure and immune-related measures have been cited selectively in expert materials, with differing emphasis on immunologic endpoints versus clinical outcomes.
Defense-side experts engage in parallel selective citation, emphasizing studies with negative findings while downplaying methodological limitations that might explain the absence of observed effects. Industry expert reports frequently cite cross-sectional studies as evidence against causation without acknowledging that such study designs have limited power to detect effects of chemical exposures that may require years or decades to manifest as clinical disease.
EPA’s health advisories for PFOA and PFOS draw primarily from the C8 Health Studies and occupational cohorts while giving minimal consideration to case-control studies and cross-sectional analyses that comprise the majority of available evidence.
Systematic reviews of PFAS and thyroid-related outcomes have reported mixed results, including positive, null, and inverse associations across studies.
Expert reports typically cite only the subset of studies that support their preferred conclusion without explaining why researchers should discount the other studies.
Dose-Response Relationships and Biological Plausibility
Establishing causation requires demonstrating that higher exposures produce greater effects, but dose-response relationships in PFAS literature are often weak or absent even in studies that report positive associations. Studies frequently find health effects only at the highest exposure category, with no gradient across lower exposure levels. This pattern is more consistent with chance findings or residual confounding than with true causal relationships.
The C8 Health Study’s findings for kidney cancer exemplify this problem. The study found increased cancer risk only in the highest quartile of exposure, with no evidence of increased risk at lower exposure levels that still exceeded background concentrations by substantial margins.
Biological plausibility varies substantially across the health outcomes commonly alleged in PFAS litigation. The evidence for PFAS associations with cholesterol levels and certain immune function measures is comparatively stronger than for many other endpoints, with findings reported across multiple study types and plausible biological pathways. The evidence for carcinogenic effects is much weaker, with inconsistent findings across studies and limited understanding of potential mechanisms.
Animal studies provide some support for PFAS carcinogenicity, but the tumors observed in rodent studies typically occur in organs and pathways that differ substantially from the cancer sites reported in human studies. Rats exposed to PFOA develop liver tumors and Leydig cell tumors in the testes, but human studies have not found consistent associations with liver cancer, and testicular cancer in humans primarily involves germ cells rather than Leydig cells.
Methodological Challenges to Causal Inference
PFAS exposure correlates with factors that independently influence disease risk: income, education, occupation, geographic location, and access to healthcare. People with higher PFAS exposures often live in areas with multiple environmental hazards, work in industrial occupations with other chemical exposures, and have different baseline disease risks than unexposed populations.
Statistical adjustment for measured confounders can introduce bias if those variables are intermediate steps in the causal pathway from exposure to disease. Studies frequently adjust for cholesterol levels when examining cardiovascular outcomes, but if PFAS exposure causes cardiovascular disease partly through effects on cholesterol, this adjustment removes rather than controls for confounding.
Reverse causation presents another concern: the possibility that disease influences exposure rather than exposure causing disease. Kidney disease, liver disease, and some cancers can alter PFAS elimination rates, potentially creating spurious associations between serum PFAS levels and disease status in studies that measure PFAS levels after disease diagnosis.
Most cancers take 10-30 years to develop after initial exposure, but few studies have sufficient follow-up time to capture this latency period. Studies with shorter follow-up may miss true associations, while studies that find associations with recent exposures may detect effects of pre-existing disease on exposure patterns rather than effects of exposure on disease risk.
Replication and Consistency Across Studies
Reproducibility of findings across independent studies provides crucial evidence for causal relationships, but PFAS health research shows limited replication of specific associations. The most consistent findings involve effects on cholesterol levels and some immune function measures, with positive associations reported across multiple populations and study designs.
Replication has been poor for cancer outcomes. Studies of kidney cancer show positive associations in some populations but not others, with no clear pattern based on exposure levels, population characteristics, or study methodology. In some topic areas, larger or more methodologically rigorous studies have reported weaker associations than smaller studies, though patterns vary by endpoint and population.
Meta-analyses attempting to synthesize findings across studies face substantial challenges due to differences in exposure assessment methods, outcome definitions, and population characteristics. Published meta-analyses typically combine studies with incompatible methodologies, producing summary estimates that may not reflect what any individual study actually demonstrated.
The geographic concentration of certain findings raises questions about generalizability to other exposure scenarios and populations.
Most studies showing strong associations between PFAS exposure and adverse health outcomes come from a small number of highly contaminated communities, particularly the Ohio River Valley region affected by DuPont’s Washington Works facility. Studies of other exposed populations, including workers at other manufacturing facilities and residents of communities with different contamination sources, have generally found weaker or absent associations.
Methodological Challenge Requirements
A rigorous methodological challenge to causation claims would focus on whether studies actually support the specific causal claims being made, rather than whether they found statistically significant associations.
Temporal relationships would receive scrutiny, particularly in studies that measured exposure after disease diagnosis or relied on cross-sectional designs. The challenge would examine whether exposure assessment methods were sufficient to detect the dose-response relationships that must exist if the alleged causal relationships are real. Claims about cancer causation would be examined against the substantial differences between tumor types observed in animal studies and those alleged in human litigation.
Statistical methodology would be assessed for multiple comparison problems, subgroup analyses performed without prior hypotheses, and the appropriateness of covariates included in statistical models. Effect sizes would be evaluated for consistency with true causal relationships versus chance findings and residual confounding.
The consistency of findings across studies would be evaluated honestly, acknowledging both positive and negative results rather than selectively citing supportive evidence. The geographic and temporal clustering of positive findings would be examined as potential evidence against causation rather than for it.
[Back]