Opinion: Why Policymakers Should Read Past The Abstract

By Manny Teodoro, Chad Seidel, and Joe Cotruvo

When it comes to drinking water, sound public policy requires sound scientific research. Publication in a prestigious, peer-reviewed journal helps establish legitimacy for scientific claims in public discourse. But science is a social process, scientific standards of evidence vary across disciplines, and peer review does not guarantee validity. For readers who stop at the abstract, these distinctions can be easy to miss.

Over the past two years, PFAS in drinking water has been the subject of major federal rulemaking, extensive public comment, and significant technical debate. As the Environmental Protection Agency crafted PFAS regulations for drinking water, the agency’s Science Advisory Board questioned whether the available evidence supports setting PFAS limits at very low concentrations. The millions of lives and billions of ratepayer dollars at stake demand careful evaluation of studies that link PFAS in drinking water to health outcomes.

It is in this context that recent claims of a new paper published in the Proceedings of the National Academy of Sciences warrant close scrutiny. The article claims that “PFAS-contaminated drinking water harms infants” based on analysis of health records in New Hampshire. The claim is striking. A close look at the evidence gives reason to question it.

The study does not demonstrate PFAS exposure empirically. The authors never establish whether pregnant women or infants in the study actually consumed drinking water with elevated PFAS concentrations. There are no serum, cord-blood, or other biomonitoring data. The analysis includes no measurements of PFAS in finished drinking water delivered to homes. It does not examine how long mothers lived at a given address, where they lived earlier in pregnancy, how much tap water they drank, or whether their water was treated or filtered. Other sources of exposure were not considered. For example, PFOS was a widely used furniture and fabric treatment which were the major source in homes. National blood levels have declined by more than 90% since that use was eliminated. These are standard expectations in toxicological research. 

Instead, the study infers PFAS exposure from distance and location relative to a contaminated site and modeled groundwater flow, then uses econometric modeling to estimate effects on infant health. The authors assume that residents “downstream” or “downgradient” from contamination sites ingested higher levels of PFAS — and that PFAS alone explains any observed health differences — despite the fact that such sites may emit multiple contaminants that can independently affect health. They also assume that residence downgradient from a contaminated groundwater site is comparable to random assignment in a laboratory for inferential purposes. 

Those assumptions matter. In reality, what comes out of a household tap reflects water treatment processes, water blending, operational decisions by utilities, and household behavior, including filtration and bottled-water use. National monitoring conducted under the U.S. Environmental Protection Agency’s Fifth Unregulated Contaminant Monitoring Program (UCMR5) measures PFAS directly in finished drinking water. These data show that when PFAS are detected in public drinking water in New Hampshire and throughout the country, concentrations are typically very low — often in the single digits of parts per trillion.

Those measured levels bear little resemblance to the exposure implied by the PNAS paper, which analyzes sites with groundwater concentrations exceeding 1,000 parts per trillion — levels associated with major sources such as industrial facilities, landfills, or firefighting departments. Although such concentrations are plausible for contaminated groundwater near a PFAS source, they are rare in finished public drinking water. Moreover, sites emitting such high PFAS concentrations are also likely sources of other contaminants; health effects attributed in the paper to PFAS alone could plausibly reflect any number of unobserved co-contaminants.

A second major concern relates to the alleged harm to infants. Infant mortality is a rare outcome. The analysis relies on records of just 56 infant deaths, and fewer than 30 “downgradient” infant deaths, from a ten-year dataset of over 100,000 births. Factors such as maternal clinical risk, exposure to other contaminants, or PFAS exposure timing and duration were not measured, but assumed away based on presumed randomness between “upgradient” and “downgradient” locations. With so few cases, misclassification of even a handful of infants could materially change the results. Well-known limitations of state birth-registry data further complicate inferences based on such small numbers.

These issues challenge the authors’ claim that they have shown PFAS exposure causes worse birth outcomes and that “PFAS-contaminated drinking water harms infants.” This is not a case of careless mathematics; the statistical work in the PNAS paper is intricate, polished, and subject to repeated robustness tests. But econometric sophistication cannot substitute for valid exposure data or established practices in toxicology, epidemiology, and drinking-water engineering. 

None of this diminishes legitimate concern about PFAS. But in the public policy arena, fragile assumptions and data limitations buried deep in technical appendices harden into authoritative claims once they are elevated by splashy titles and provocative abstracts. Those claims then migrate rapidly into headlines, advocacy campaigns, and policy debates, often without the scrutiny that their underlying evidence warrants.

Protecting public health requires vigilance, but it also requires scientific discipline. As environmental risks grow more complex, policymakers protect the public not by reacting to headlines or conclusions alone, but by insisting on evidence that demonstrates exposure, reflects real-world conditions, and meets the standards required for consequential decisions.

About the authors:

Manny Teodoro is a professor at the La Follette School of Public Affairs at UW-Madison. 

Chad Seidel, PhD, is a professional drinking water engineer with more than 20 years of experience serving the drinking water community. 

Joseph Cotruvo is a former director of EPA’s Drinking Water Standards Division and a water-quality scientist specializing in risk assessment and drinking-water treatment.