Retracted Iowa Water Study Shows How Misclassified Samples Can Distort Nitrate Risk Findings

A study that linked prenatal nitrate exposure in Iowa’s public drinking water to birth outcomes has been retracted, after its own author discovered that roughly one-third of the water samples in the analysis were not public drinking water at all. The retraction notice, published by the editors of PLOS Water on August 31, 2026, formally withdraws the 2025 paper, which had examined early prenatal nitrate exposure and birth outcomes using Iowa public drinking water records from 1970 to 1988. The case offers an unusually transparent look at how a single data-classification error, compounded by statistical and exposure-timing problems, can undermine the validity of an environmental health finding.

The sequence of events began shortly after the original article appeared. The author, who had published the study in PLOS Water in 2025, contacted the journal to report a serious problem with the underlying data. Through the University of Iowa’s Center for Health Effects of Environmental Contamination, the author learned that approximately one-third of the analyzed samples had been misclassified: they were not samples of treated public drinking water, as the study assumed, but untreated ground or surface water instead. That distinction matters enormously in this field, because public water systems are subject to regulatory monitoring and treatment requirements that untreated sources are not, and nitrate concentrations in the two types of water can differ substantially.

The misclassification struck at the heart of the study’s central claim. The original paper reported an association between birth weight and nitrate levels in public drinking water, a finding that would have carried implications for how pregnant people and health agencies think about agricultural nitrate runoff, a persistent contamination problem across Iowa and much of the Corn Belt. If a sizable fraction of the samples labeled as public drinking water actually came from untreated wells or surface sources, then the exposure estimates assigned to each birth no longer described what the study claimed they described. When the author repeated the analysis after excluding all non-drinking-water samples, the results changed in meaningful ways, confirming that the error was not a trivial technicality but a driver of the published findings.

As is standard practice, a member of the PLOS Water Editorial Board assessed the author’s concern, and the review surfaced additional problems beyond the sample misclassification. The first involved multiple testing. The original analysis had conducted multiple statistical tests, but reported no methods to correct for multiple testing, a step that researchers use to reduce the risk of false positive associations. When many tests are run across different exposure windows, outcomes, and subgroups, some will appear statistically significant by chance alone. Without an explicit correction, such as adjustments to significance thresholds or formal procedures that account for the number of comparisons, a reported association becomes difficult to distinguish from statistical noise.

The second additional concern concerned the temporal logic of exposure assessment. The study used a 30-day window to link each birth to a nitrate measurement, intending to capture exposure around the time of conception. The editorial assessment noted that this window may not accurately represent true exposure at conception, because nitrate concentrations in a water supply can shift substantially within a span of 30 days. Seasonal agricultural runoff, rainfall, and changes in pumping can all move nitrate levels up or down in a matter of weeks, meaning a single measurement may poorly represent the water a person actually drank during the relevant period of early pregnancy.

There is also a subtler problem the editors flagged: remediation and source switching can invert the expected exposure-response pattern. When a public water system exceeds the nitrate standard, mandatory remediation kicks in, or the system shifts to a different water source, and nitrate concentrations can drop quickly. If systems with historically high nitrate are precisely the ones that remediated and now deliver low-nitrate water, then a study linking births to measured nitrate levels may find that communities with a history of contamination show lower current exposure than expected. Depending on how the data are structured, that dynamic can scramble or even reverse the apparent relationship between nitrate and birth outcomes, producing a pattern that looks like a finding but is actually an artifact of the regulatory response itself.

In light of these concerns, which the editors concluded called into question the validity of the published results, the PLOS Water Editors retracted the article. Importantly, the author agreed with the retraction. That agreement is worth noting: retractions are often contested, adversarial processes, but in this case the author’s own vigilance in identifying the data problem and reporting it to the journal set the correction in motion. The retraction notice cites the original paper, a 2025 study in PLOS Water, and the retraction notice itself carries the canonical DOI for the withdrawal record in the journal’s fifth volume.

The scientific stakes in this case are real. Nitrate contamination of drinking water is one of the most widespread agricultural pollution problems in the United States, and Iowa sits at its center, with intensive row cropping and livestock production feeding nitrogen into the aquifers and rivers that supply public water systems. The federal maximum contaminant level for nitrate in public water supplies is set to protect against acute toxicity, particularly methemoglobinemia in infants, but researchers have long investigated whether exposure at or below the regulatory limit during pregnancy is associated with subtler outcomes such as birth weight and preterm birth. Studies in this area typically rely on historical monitoring records linked to birth certificates, exactly the kind of data infrastructure the retracted study used. Because the exposure is diffuse and the outcomes are influenced by many social and medical factors, the field depends heavily on careful exposure classification and conservative statistical practice; errors in either direction can ripple through policy debates about water treatment costs and agricultural regulation.

The retraction also illustrates how the modern correction ecosystem is supposed to work, and where it remains fragile. The author’s discovery came not through peer review, which had already approved the paper, but through follow-up with the University of Iowa’s Center for Health Effects of Environmental Contamination, an institution that maintains long-term records on water quality in the state. That kind of post-publication scrutiny is increasingly central to research integrity, since no review process can fully verify the provenance of every data point in a study spanning nearly two decades of historical records. Journals such as PLOS Water publish retraction notices that document the specific technical reasons for withdrawal, giving other researchers a detailed map of what went wrong rather than a bare announcement. The open-access nature of the record means the original paper, the retraction notice, and the reasoning behind both remain publicly visible, allowing the field to learn from the failure.

For readers who encountered the original findings, the practical takeaway is that the reported association between early prenatal nitrate exposure in Iowa’s public drinking water and birth outcomes should no longer be treated as reliable evidence. The retraction does not establish that nitrate exposure has no effect on birth weight; it establishes only that this particular analysis cannot support its original conclusion, because of the sample misclassification, the uncorrected multiple testing, and the exposure-timing concerns. Future work on the question will need to distinguish treated public water from untreated sources with precision, apply appropriate corrections when testing many associations, and model exposure in ways that account for within-month variability and the rapid changes that follow regulatory exceedances. Until such studies are done with those safeguards in place, the Iowa data from 1970 to 1988 remain what they always were: a rich but demanding record that requires exactly the kind of scrutiny this retraction demonstrates.

Subject of Research: Retraction of a study on prenatal nitrate exposure in Iowa public drinking water and birth outcomes

Article Title: Retraction: Early prenatal nitrate exposure and birth outcomes: A study of Iowa’s public drinking water (1970–1988)

Article References: The PLOS Water Editors (2026). Retraction: Early prenatal nitrate exposure and birth outcomes: A study of Iowa’s public drinking water (1970–1988). PLOS Water, 5(8), e0000622. https://doi.org/10.1371/journal.pwat.0000622

Image Credits: AI Generated

DOI: 10.1371/journal.pwat.0000622

Keywords: nitrate, drinking water, birth weight, retraction, Iowa, PLOS Water, prenatal exposure, water contamination, research integrity, public health, groundwater, statistical methods