Proposed Section Structure
This article is organized into the following sections: Highlights; Background and Clinical Context; Study Design and Methods; Key Results; Clinical and Scientific Interpretation; Strengths and Limitations; Implications for Practice and Research; Conclusion; Funding and Trial Registration; Citation.
Highlights
In a pooled analysis of 5,229 U.S. pregnancies from 16 ECHO cohorts, higher maternal concentrations of six commonly detected PFAS were not associated with a higher prevalence of gestational diabetes mellitus (GDM).
The overall PFAS mixture showed an inverse association with GDM prevalence, with a prevalence ratio of 0.75 per quartile increase (95% CI 0.58 to 0.96), a finding that should be interpreted cautiously given the biologic and epidemiologic complexity of PFAS exposure assessment in pregnancy.
Associations with fasting glucose in the subgroup with biomarker data were largely null, although some positive and negative associations appeared within specific race and ethnicity strata.
Effect modification analyses suggested stronger inverse associations for some PFAS among participants with prepregnancy BMI less than 25 kg/m2, underscoring the possibility of heterogeneity by metabolic phenotype.
Background and Clinical Context
Gestational diabetes mellitus remains a major complication of pregnancy, with important short- and long-term implications for both mother and child. For pregnant individuals, GDM increases risks of hypertensive disorders, cesarean delivery, and future type 2 diabetes. For offspring, exposure to maternal hyperglycemia is associated with fetal overgrowth, neonatal metabolic complications, and an elevated long-term risk of obesity and glucose dysregulation. Because traditional risk factors such as adiposity, age, family history, and prior GDM do not fully explain disease occurrence, attention has increasingly turned to environmental contributors.
Per- and polyfluoroalkyl substances, or PFAS, are persistent synthetic chemicals widely used in industrial processes and consumer products. Their high environmental stability has led to broad human exposure through drinking water, food packaging, household items, and occupational sources. PFAS are of clinical and public health interest because they can cross the placenta, persist in the body for years, and may interfere with lipid metabolism, endocrine signaling, inflammatory pathways, and insulin sensitivity. These properties make them plausible candidates for affecting pregnancy-related glucose homeostasis.
Yet the epidemiologic literature on PFAS and GDM has been inconsistent. Some studies have suggested increased risk, others have shown null findings, and a few have suggested inverse associations. Differences in timing of PFAS measurement, outcome definitions, sample size, confounding control, and physiologic changes of pregnancy likely contribute to these discrepancies. The ECHO analysis by Starling and colleagues is therefore notable for its scale, pooled multicohort design, and explicit mixture modeling.
Study Design and Methods
Population
The investigators analyzed 5,229 participants with singleton pregnancies drawn from 16 U.S. cohorts within the Environmental influences on Child Health Outcomes, or ECHO, Consortium. Pregnancies occurred between 1999 and 2021. This pooled design improved statistical power and increased geographic and demographic diversity compared with many earlier studies.
Exposure Assessment
PFAS were measured in one maternal plasma or serum sample collected during pregnancy. Six PFAS detected in at least 60% of participants were included in the primary analyses. The abstract does not list all six analytes, but the selection criterion suggests a focus on compounds with sufficiently widespread detection to support robust modeling.
A key methodological issue in studies of PFAS during pregnancy is that measured concentrations may reflect not only external exposure but also pregnancy physiology. Plasma volume expansion, altered protein binding, kidney filtration changes, and parity-related transfer from mother to fetus or via prior lactation can all influence measured levels. These factors are important when interpreting inverse or null associations.
Outcomes
The primary outcome was GDM, ascertained by self-report or medical record documentation. The use of both sources may improve case capture but can also introduce some heterogeneity, especially if diagnostic criteria varied across cohorts and calendar years.
The secondary outcome was fasting glucose, available in a subgroup of 1,213 participants. Fasting glucose offers a continuous metabolic endpoint that may detect subtler glycemic variation than a binary GDM diagnosis, although it does not fully capture post-load glucose abnormalities.
Statistical Approach
For GDM, the investigators used generalized estimating equation models with a Poisson distribution and robust variance to estimate prevalence ratios. This is an appropriate approach for binary outcomes when relative measures are preferred over odds ratios, especially in pooled cohort analyses. For fasting glucose, generalized estimating equation linear regression models were used.
The authors also evaluated effect modification by prepregnancy BMI and by race and ethnicity using interaction terms and stratified analyses. To address the reality that PFAS occur as correlated mixtures rather than isolated exposures, they applied quantile-based g-computation to estimate the joint effect of the six-PFAS mixture.
Key Results
Overall Association With GDM
The principal finding was straightforward: greater maternal PFAS concentrations were not associated with a higher prevalence of gestational diabetes. Across individual PFAS, associations were null or weakly inverse. This finding directly counters the concern that PFAS exposure, at least as captured in this pooled U.S. dataset, is a major independent driver of GDM risk.
The mixture analysis went a step further and showed a statistically significant negative association between the six-PFAS mixture and GDM. The prevalence ratio per quartile increase in the mixture was 0.75, with a 95% confidence interval of 0.58 to 0.96. On its face, this suggests lower GDM prevalence with higher combined PFAS levels. However, in environmental epidemiology, inverse associations involving persistent chemicals in pregnancy require careful interpretation rather than a simple conclusion of benefit. Reverse causation is unlikely in the strict temporal sense, but pregnancy-related physiology, residual confounding, and outcome misclassification could all influence the observed direction.
Fasting Glucose Findings
Among the 1,213 participants with fasting glucose data, associations were largely null. This is clinically important because if PFAS were materially worsening glucose regulation during pregnancy, one might expect at least a consistent upward shift in fasting glucose. That pattern was not observed.
At the same time, the authors reported both positive and negative associations in specific race and ethnicity strata. These subgroup findings are hypothesis-generating rather than definitive. They may reflect biologic heterogeneity, differences in exposure sources, variations in social or structural determinants that correlate with both exposure and metabolic health, or chance findings from multiple comparisons.
Effect Modification by Prepregnancy BMI
Some PFAS showed more strongly negative associations with GDM among participants with prepregnancy BMI below 25 kg/m2. This interaction is intriguing. BMI is a dominant risk factor for GDM, and among individuals with lower baseline metabolic risk, smaller physiologic or methodological shifts may become more visible. Conversely, in those with overweight or obesity, stronger conventional risk pathways may overshadow weaker environmental effects. Another possibility is that PFAS pharmacokinetics differ by adiposity or related metabolic traits.
Clinical and Scientific Interpretation
The central message for clinicians is that this large U.S. multicohort study does not support a clinically meaningful increase in GDM prevalence associated with higher prenatal PFAS concentrations. That does not mean PFAS are harmless in pregnancy. Rather, it means that within the exposure ranges and methods captured here, the data do not implicate PFAS as a clear GDM risk factor.
This distinction matters. PFAS have been associated in other contexts with dyslipidemia, immune effects, altered liver enzymes, and developmental concerns. The absence of a strong PFAS-GDM signal should not be generalized to all maternal-fetal outcomes. It should instead narrow the field: PFAS may not be a major contributor to gestational dysglycemia, or current epidemiologic tools may still be insufficiently precise to detect a modest effect.
Why might the observed associations be null or inverse? Several explanations are plausible. First, a single PFAS measurement during pregnancy may inadequately represent etiologically relevant exposure, particularly if the critical window for metabolic disruption is preconception or early first trimester. Second, GDM diagnosis is not uniform across place and time; varying screening practices and thresholds may dilute true associations. Third, physiologic changes of pregnancy can alter serum concentrations of persistent chemicals in ways that correlate with metabolic status. For example, hemodilution or changes in renal handling could complicate interpretation of biomarker levels. Fourth, PFAS are a heterogeneous chemical class, and effects may vary by compound rather than move in the same direction across a mixture.
The inverse mixture finding is perhaps the study’s most provocative result. Clinically, it should not be interpreted as evidence that PFAS protect against GDM. More likely, it reflects one or more of the methodological issues above, or residual confounding by factors associated with both PFAS exposure and healthcare access, diet, parity, or sociodemographic characteristics. In environmental epidemiology, a statistically significant inverse association is not automatically causal in the protective direction.
Strengths and Limitations
Strengths
The study has several clear strengths. Its sample size of 5,229 pregnancies gives it substantially greater precision than many prior PFAS-GDM studies. The inclusion of 16 cohorts increases diversity and enhances generalizability within the U.S. context. The use of both single-pollutant and mixture models is methodologically important because PFAS exposures are correlated in real life. Evaluation of fasting glucose as a secondary continuous outcome adds metabolic depth beyond a binary diagnosis. Finally, assessment of effect modification by BMI and race and ethnicity acknowledges that environmental exposures do not operate uniformly across populations.
Limitations
Several limitations temper interpretation. PFAS were measured once during pregnancy, which may miss exposure variability and obscure the most relevant susceptibility window. GDM ascertainment relied on self-report or medical records, and the abstract does not specify whether diagnostic criteria were harmonized across cohorts. Residual confounding remains possible, especially for diet, parity, kidney function, socioeconomic factors, and co-exposures. Only 1,213 participants had fasting glucose data, reducing power for that endpoint. In addition, subgroup findings by BMI and race and ethnicity should be interpreted cautiously because stratified analyses increase the chance of unstable estimates and false-positive results.
Another important limitation is the challenge of translating serum or plasma PFAS levels measured during pregnancy into causal inferences about metabolic disease. Pregnancy is a dynamic physiologic state, and biomarker concentrations can be influenced by the very metabolic processes under investigation. This is a longstanding issue in studies of persistent organic pollutants and pregnancy outcomes.
Implications for Practice and Research
For obstetric and endocrine practice, these results do not support incorporating PFAS biomarker testing into routine GDM risk assessment. Established screening strategies should remain anchored in known clinical risk factors and evidence-based diagnostic pathways. Patients who ask about environmental chemicals should receive balanced counseling: reducing avoidable PFAS exposure remains reasonable from a precautionary public health standpoint, but current evidence does not indicate that higher PFAS exposure clearly raises GDM risk.
For researchers, the study highlights several priorities. First, future work should examine preconception and early-pregnancy exposure windows, ideally with repeated PFAS measures. Second, harmonized and preferably laboratory-based outcome definitions for GDM are essential. Third, studies should better account for physiologic determinants of PFAS concentrations during pregnancy, including hemodilution, renal function, and parity. Fourth, mixture methods remain crucial, but compound-specific analyses should continue because PFAS are not toxicologically interchangeable. Finally, the subgroup signals by BMI and race and ethnicity deserve follow-up in studies designed to evaluate heterogeneity rather than merely explore it.
More broadly, this study is a reminder that environmental risk factor research must meet a high methodological bar before informing clinical policy. Large pooled cohorts are valuable, but exposure timing, outcome harmonization, and biologic interpretation are equally critical. Null findings are important when they come from strong datasets; they refine the field and prevent overstatement of uncertain hazards.
Conclusion
In this large pooled analysis from the ECHO Cohort, maternal PFAS concentrations during pregnancy were not associated with a higher prevalence of gestational diabetes, and fasting glucose associations were mostly null. Although the PFAS mixture showed an inverse association with GDM, that result should be interpreted cautiously rather than as evidence of benefit. For clinicians, the study reduces concern that PFAS are a major direct driver of GDM in contemporary U.S. populations, while still leaving open important questions about exposure timing, subgroup susceptibility, and other maternal-fetal outcomes.
Funding and Trial Registration
The provided abstract does not specify funding details. The study was conducted within the Environmental influences on Child Health Outcomes (ECHO) Cohort Consortium. No ClinicalTrials.gov registration number is provided in the supplied citation information.
Citation
Starling AP, Burjak M, Nzegwu AW, Xun X, Adgate JL, Barrett ES, Bennett DH, Chatzi L, Colicino E, Dabelea D, Dunlop AL, Eick SM, Farzan SF, Ferrara A, Fleisch AF, Geiger SD, Hedderson MM, Kahn LG, Karagas MR, Kelly RS, Liang D, Lin PI, O’Connor TG, Padula AM, Peterson AK, Romano ME, Sathyanarayana S, Zhu Y, Valvi D, ECHO Cohort Consortium. Per- and Polyfluoroalkyl Substances During Pregnancy and Gestational Diabetes: The Environmental Influences on Child Health Outcomes (ECHO) Cohort. Diabetes Care. 2026 May 1;49(5):852-860. PMID: 41875060. URL: https://pubmed.ncbi.nlm.nih.gov/41875060/
