Associations of Perfluoroalkyl and Polyfluoroalkyl Substances With Cardiovascular Disease Incidence in Adults With Prediabetes: Findings From the Diabetes Prevention Program

Abstract

Background

Perfluoroalkyl and polyfluoroalkyl substances (PFAS) are persistent, widespread environmental contaminants linked to cardiometabolic outcomes including obesity, hyperlipidemia, and diabetes. We examined whether baseline plasma PFAS concentrations are associated with incident cardiovascular disease (CVD) in adults with prediabetes, leveraging data from DPPOS (Diabetes Prevention Program Outcomes Study).

Methods

Among 1382 participants, we quantified baseline plasma concentrations of 6 PFAS. We used Cox proportional hazards models to estimate the risks of developing CVD outcomes during a median of 21 years of follow‐up for each PFAS and used quantile g‐computation to evaluate the joint effect of all 6 PFAS. Effect modification by age, sex, menopausal status, diet, and physical activity was explored.

Results

The incidence of major adverse cardiovascular events was 9.6%; 3.9% had CVD‐related death. Each increase in interquartile range (1.1 ng/mL) in 2‐(N‐methyl‐perfluorooctane sulfonamido) acetate was associated with a 16% higher risk of major adverse cardiovascular events (95% CI, 1–33%) and a 24% higher risk of CVD death (95% CI, 2–52%). Higher concentrations of perfluorohexane sulfonate, perfluorooctane sulfonate, 2‐(N‐ethyl‐perfluorooctane sulfonamido) acetate, and perfluorooctanoate were associated with greater risk of CVD outcomes, including nonfatal myocardial infarction, hospitalized congestive heart failure, and cardiovascular death. However, PFAS mixture was not associated with CVD. Age, sex, treatment arm, physical activity, and diet did not modify the associations of individual PFAS.

Conclusion

In adults with prediabetes, higher plasma concentrations of select PFAS, but not their mixture, were prospectively associated with increased CVD risk. These findings underscore PFAS as a potential environmental risk factor for CVD in high‐risk populations.

REGISTRATION: URL: https://clinicaltrials.gov/; Unique identifiers: NCT00004992 and NCT00038727.

Nonstandard Abbreviations and Acronyms

CDC

Centers for Disease Control and Prevention

DPP

Diabetes Prevention Program

DPPOS

Diabetes Prevention Program Outcomes Study

EtFOSAA

2‐(N‐ethyl‐perfluorooctane sulfonamido) acetate

extended MACE

major adverse cardiovascular events plus hospitalization for congestive heart failure or unstable angina, coronary/peripheral revascularization, angiography‐confirmed coronary disease, or silent myocardial infarction

HDI

Healthy Diet Index

MACE

major adverse cardiovascular events

MeFOSAA

2‐(N‐methyl‐perfluorooctane sulfonamido) acetate

NHANES

National Health and Nutrition Examination Survey

NIDDK

National Institute of Diabetes and Digestive and Kidney Diseases

PFAS

perfluoroalkyl and polyfluoroalkyl substances

PFOA

perfluorooctanoate

PFHxS

perfluorohexane sulfonate

PFNA

perfluorononanoate

PFOS

perfluorooctane sulfonate

Clinical Perspective.

What Is New?

• In high‐risk prediabetic populations, specific perfluoroalkyl and polyfluoroalkyl substances (notably 2‐[N‐methyl‐perfluorooctane sulfonamido] acetate) were associated with increased cardiovascular disease risk, while their combined mixture was not.

• These associations were consistent among different age, sex, and lifestyle subgroups, indicating that perfluoroalkyl and polyfluoroalkyl substances represent a broad risk factor.

What Are the Clinical Implications?

• Exposure to specific perfluoroalkyl and polyfluoroalkyl substances should be considered an emerging environmental risk factor for cardiovascular disease, particularly in metabolically vulnerable patients.

Perfluoroalkyl and polyfluoroalkyl substances (PFAS) are a group of >12 000 synthetic chemicals that have become widespread environmental and human contaminants. ¹ Their extensive use in consumer products (eg, stain‐resistant fabrics, nonstick cookware, food packaging) and industrial applications has led to the global population’s exposure, affecting communities worldwide. ² Colloquially known as “forever chemicals” due to their stability, some PFAS have elimination half‐lives of several years in humans, leading to long‐term bioaccumulation and potential health harms. ³ , ⁴

Growing evidence suggests that PFAS disrupt cardiometabolic health through multiple pathways. For example, PFAS can alter lipid metabolism, elevating low‐density lipoprotein cholesterol cholesterol, which may accelerate atherosclerosis. ² , ⁵ PFAS can also trigger chronic inflammation, oxidative stress, and endothelial dysfunction, which are key drivers of cardiovascular disease (CVD). ⁶ , ⁷ , ⁸ , ⁹ Other studies show that PFAS can impair vascular function by increasing endothelial permeability and promoting hypercoagulability through platelet membrane incorporation, thereby contributing to hypertension, arterial stiffness, and heightened risks of thrombosis, myocardial infarction (MI), and stroke. ¹⁰ , ¹¹

Despite mechanistic plausibility, epidemiological evidence on the association between PFAS and incident CVD remains limited and inconsistent. ⁹ The National Academies of Sciences, Engineering, and Medicine noted in their 2022 clinical guidelines that there are inadequate or insufficient data on hard clinical end points, such as MI or heart failure, to draw robust conclusions. ¹² While some analyses using data from the National Health and Nutrition Examination Survey (NHANES) have reported positive associations between PFAS and self‐reported CVD risk, ¹³ , ¹⁴ these findings have not been consistently replicated in other cohorts or in a recent meta‐analysis. ¹⁵ This discrepancy may arise from methodological limitations, including cross‐sectional design of many studies, residual confounding, and heterogeneity in PFAS exposure profiles among populations. In addition, differences in outcome adjudication highlight the need for studies with rigorous CVD phenotyping.

The DPP (Diabetes Prevention Program) and DPPOS (Diabetes Prevention Program Outcomes Study) provides a valuable opportunity to address these gaps. ¹⁶ With over 20 years of follow‐up in a well‐characterized cohort of participants with prediabetes, DPP/DPPOS offers PFAS exposure assessment data, rigorously adjudicated CVD outcomes, and data on key potential effect modifiers such as age, sex, and lifestyle factors, including diet and physical activity. Our prior analyses of DPP/DPPOS linked PFAS to hyperlipidemia ¹⁷ and subclinical atherosclerosis such as calcifications of the coronary and aortic arteries. ¹⁸ In the present study, we leveraged DPP/DPPOS data to prospectively evaluate associations between plasma PFAS concentrations and incident CVD outcomes during 21 years of median follow‐up. By analyzing PFAS both individually and as a mixture, we aimed to clarify whether specific PFAS or their combined burden drive CVD risk in metabolically vulnerable populations. We hypothesized that greater PFAS plasma concentrations would be positively associated with increased incidence of CVD events, with stronger effects in women and older individuals, and that healthy lifestyle behaviors would mitigate risk.

Methods

Study Design and Population

DPP was a landmark multicenter randomized controlled trial (recruitment 1996–1999) that evaluated interventions to prevent type 2 diabetes in high‐risk adults (aged ≥25 years) with overweight/obesity and impaired fasting glucose. ¹⁹ The data used for this study are available from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Central Repository but restrictions apply. These data were used under a data use agreement for the current study and are not publicly available. However, data are available from the authors on reasonable request and with permission of the NIDDK.

Participants were randomized to: (1) an intensive lifestyle intervention targeting 7% weight loss through 150 minutes per week of moderate exercise and a healthy low‐calorie, low‐fat diet; (2) metformin; or (3) placebo. ²⁰ After DPP concluded in 2002, participants transitioned to DPPOS, which implemented a modified lifestyle intervention among all arms and provided open‐label metformin to the original metformin group during extended follow‐up to assess long‐term intervention effects.

For this analysis, we used the DPP February 2008 Full Scale Data Release and April 2024 DPPOS Phase 3 Data Release available from the NIDDK repository. We measured plasma PFAS concentrations at DPP randomization to evaluate the effect of PFAS on the first occurrence of CVD events during 21 years of median follow‐up. From 3081 DPP participants randomized to lifestyle, metformin, or placebo arms, we excluded 26 who did not consent to release CVD data to the NIDDK repository and 1673 who lacked sufficient plasma for quantifying PFAS, resulting in 1382 with both CVD and PFAS measurements for final analysis (See Figure S1 for the study flow chart). All participants in DPP/DPPOS provided written informed consent. The current study received approval from the institutional review board at Harvard Pilgrim Health Care. The involvement of the Centers for Disease Control and Prevention (CDC) laboratory did not constitute engagement in human subject research.

Plasma PFAS Concentrations

We obtained archived plasma samples collected from DPP participants at baseline (1996–1999) from the NIDDK repository and analyzed them at the CDC laboratory using online solid‐phase extraction high‐performance liquid chromatography‐isotope dilution‐tandem mass spectrometry. ²¹ , ²² The limit of detection was 0.1 ng/mL for all PFAS. For concentrations below the limit of detection LOD, we applied the limit of detection/√2 imputation method following established protocols. ²³ As we did in previous analyses, ¹⁷ we summed branched and linear isomers of perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) to calculate total concentrations of PFOS and PFOA, respectively. The CDC followed established quality assurance programs, in compliance with the requirements set forth in the Clinical Laboratory Improvement Amendments of 1988.

We a priori included PFAS with detectable concentrations in >60% of samples in the final statistical analyses: 2‐(N‐methyl‐perfluorooctane sulfonamido) acetate (MeFOSAA), 2‐(N‐ethyl‐perfluorooctane sulfonamido) acetate (EtFOSAA), perfluorohexane sulfonate (PFHxS), PFOS, PFOA, and perfluorononanoate (PFNA). MeFOSAA and EtFOSAA notably eventually metabolized to PFOS. ²⁴ We measured PFAS concentrations in placebo (n=472) and lifestyle arm (n=478) samples in 2018, followed by metformin arm samples in 2024 (n=432). Prior analyses comparing repeated measurements among different timepoints confirmed no laboratory drift. ²⁵ The 2024 analytical panel excluded EtFOSAA based on its removal from CDC’s standard PFAS protocol. We calculated the PFAS Burden Score using the method developed by Liu et al. ²⁶ , ²⁷ This standardized metric integrates the most frequently detected PFAS, expressed as SDs relative to the NHANES 2017 to 2018 population mean, enabling cross‐study comparability. More details of the PFAS analytical methods can be found in our prior publication, which also showed that the distribution of PFAS concentrations in DPP/DPPOS participants were comparable with the NHANES population from a similar period. ³

CVD Incidence

DPP/DPPOS screened for CVD events at every study visit. Detailed methods for CVD outcome ascertainment have been previously described. ²⁸ Briefly, an independent Outcomes Classification Committee of blinded physicians with consistent membership throughout the study adjudicated all events using standardized criteria, reviewing comprehensive medical documentation. In this analysis, we included adjudicated cardiovascular events available in the NIDDK repository and evaluated the first occurrence of a major adverse cardiovascular event (MACE; comprising nonfatal MI [excluding silent MI], nonfatal stroke, or CVD death), extended MACE (MACE plus hospitalization for congestive heart failure or unstable angina, coronary/peripheral revascularization, angiography‐confirmed coronary disease, or silent MI), and individual end points with sufficient sample size, including nonfatal MI, nonfatal stroke, congestive heart failure hospitalization, coronary revascularization, and CVD death.

Covariates/Effect Modifiers

We selected covariates a priori based on our study hypotheses and employed directed acyclic graphs to identify potential confounders while avoiding adjustment for mediators on the causal pathway between plasma PFAS concentrations and CVD outcomes (Figure S2). We adjusted models for the following baseline covariates in the main analysis: age, sex and menopausal status (for women), race and ethnicity, treatment assignment, household income, diet quality, and smoking status (Table 1). In sensitivity analyses, we evaluated the following additional variables: educational attainment, marital status, body mass index (calculated as weight in kilograms divided by height in meters squared), physical activity, alcohol consumption, and fish intake. We extracted covariates directly from the NIDDK repository, with all data collected using standardized DPP/DPPOS study forms. The DPP collected information on usual diet using a semiquantitative food frequency questionnaire (FFQ) ²⁹ and assessed diet quality using the Healthy Diet Index (HDI), ³⁰ which is a 9‐point composite score incorporating 4 nutrients (saturated fatty acids, polyunsaturated fatty acids, monosaccharides and disaccharides, and cholesterol) and 5 food groups (complex carbohydrates, dietary fiber, fruits and vegetables, and pulses/nuts/seeds). DPP assessed physical activity level using the NHANES III Physical Activity Scale, ²¹ which asked participants about the frequency of leisure‐time activities (eg, walking, running, cycling, swimming, aerobics, dancing, calisthenics, gardening, weightlifting, other activities) in the past month. There were no missing data for any covariates among included participants.

Table 1.

Baseline Demographic Characteristics (1996–1999) and CVD Incidence by Tertile^* of Baseline Plasma PFAS Burden Score Among 1382 Participants From DPP and DPPOS

	All (N=1382)	PFAS tertile 1 (n=463)	PFAS tertile 2 (n=457)	PFAS tertile 3 (n=462)	P value
Women	900 (65.1%)	367 (79.3%)	281 (61.5%)	252 (54.5%)	<0.01
Treatment group
Lifestyle	478 (34.6%)	170 (36.7%)	155 (33.9%)	153 (33.1%)	0.01
Metformin	432 (31.3%)	119 (25.7%)	142 (31.1%)	171 (37.0%)
Placebo	472 (34.2%)	174 (37.6%)	160 (35.0%)	138 (29.9%)
Age at baseline, mean±SD, y	52.1 (10.3)	50.3 (10.2)	52.5 (10.3)	53.5 (10.0)	<0.01
Race and ethnicity
White	818 (59.2%)	269 (58.1%)	284 (62.1%)	265 (57.4%)	<0.01
Black	278 (20.1%)	64 (13.8%)	72 (15.8%)	142 (30.7%)
Hispanic, of any race	228 (16.5%)	102 (22.0%)	82 (17.9%)	44 (9.5%)
All other†	58 (4.2%)	28 (6.0%)	19 (4.2%)	11 (2.4%)
Menopausal status‡
Premenopausal	398 (44.2%)	200 (54.5%)	118 (42.0%)	80 (31.7%)	<0.01
Postmenopausal	502 (55.8%)	167 (45.5%)	163 (58.0%)	172 (68.3%)
Household income, $
<20 000	163 (11.8%)	73 (15.8%)	46 (10.1%)	44 (9.5%)	<0.01
20 000—34 999	233 (16.9%)	91 (19.7%)	62 (13.6%)	80 (17.3%)
35 000—49 999	271 (19.6%)	84 (18.1%)	104 (22.8%)	83 (18.0%)
50 000—74 999	283 (20.5%)	94 (20.3%)	99 (21.7%)	90 (19.5%)
≥75 000	320 (23.2%)	79 (17.1%)	117 (25.6%)	124 (26.8%)
No response	112 (8.1%)	42 (9.1%)	29 (6.3%)	41 (8.9%)
Marital status
Married or living together	929 (67.2%)	307 (66.3%)	324 (70.9%)	298 (64.5%)	0.16
Divorced or separated	227 (16.4%)	89 (19.2%)	63 (13.8%)	75 (16.2%)
Widowed	65 (4.7%)	19 (4.1%)	20 (4.4%)	26 (5.6%)
Never married	161 (11.6%)	48 (10.4%)	50 (10.9%)	63 (13.6%)
Education
Not a high school graduate	734 (53.1%)	246 (53.1%)	233 (51.0%)	255 (55.2%)	0.48
High school graduate/GED	283 (20.5%)	102 (22.0%)	91 (19.9%)	90 (19.5%)
Some college and above	365 (26.4%)	115 (24.8%)	133 (29.1%)	117 (25.3%)
Smoking status
Never	796 (57.6%)	292 (63.1%)	265 (58.0%)	239 (51.7%)	0.02
Current	76 (5.5%)	23 (5.0%)	24 (5.3%)	29 (6.3%)
Former	510 (36.9%)	148 (32.0%)	168 (36.8%)	194 (42.0%)
BMI, mean±SD	33.6 (6.52)	33.6 (6.49)	33.9 (6.72)	33.4 (6.35)	0.53
HDI§, mean±SD	2.55 (1.27)	2.58 (1.26)	2.54 (1.27)	2.52 (1.28)	0.77
CVD incidence
MACE||	132 (9.6%)	36 (7.8%)	43 (9.4%)	53 (11.5%)	0.16
Extended MACE	208 (15.1%)	57 (12.3%)	67 (14.7%)	84 (18.2%)	0.04
Nonfatal MI	60 (4.3%)	15 (3.2%)	20 (4.4%)	25 (5.4%)	0.27
Nonfatal stroke	31 (2.2%)	8 (1.7%)	8 (1.8%)	15 (3.2%)	0.20
Hospitalized CHF	45 (3.3%)	16 (3.5%)	15 (3.3%)	14 (3.0%)	0.94
Coronary revascularization	91 (6.6%)	16 (3.5%)	35 (7.7%)	40 (8.7%)	0.01
CVD death	54 (3.9%)	18 (3.9%)	17 (3.7%)	19 (4.1%)	0.95

Values are presented as number (percentage) unless otherwise indicated. BMI indicates body mass index (calculated as weight in kilograms divided by height in meters squared); CHF, congestive heart failure; CVD, cardiovascular disease; DPP, Diabetes Prevention Program; DPPOS, Diabetes Prevention Program Outcome Study; GED, General Educational Development; HDI, Healthy Diet Index; MACE, major adverse cardiovascular events; MI, myocardial infarction; and PFAS, perfluoroalkyl and polyfluoroalkyl substances.

^*PFAS Burden Score for tertile 1: −1.33 to 1.28; tertile 2: >1.28 to 1.52; and tertile 3: >1. 52 to 2.36.

^† “All other” includes non‐Hispanic Asian American/Pacific Islander (AAPI) and Other (including multiracial, Alaskan Native, and non‐Hispanic Native American).

^‡ Among female participants.

^§ The HDI in this study ranged from 0 to 7 (theoretical maximum: 9).

^|| MACE included incidence of nonfatal MI, nonfatal stroke, or CVD death. Extended MACE included MACE and hospitalized CHF, coronary or peripheral revascularization, congestive heart disease by angiography, silent MI.

Statistical Analysis

We summarized participant characteristics using descriptive statistics, assessed variable distributions via normality tests, and inspected for outliers graphically. Plasma PFAS concentrations had skewed distributions, and log‐transformation did not improve model fit according to Akaike information criterion comparisons; therefore, we followed recommendations by Choi et al ³¹ and analyzed untransformed plasma PFAS concentrations. We examined correlations between PFAS using Spearman correlation.

We used Cox proportional hazards models to estimate hazard ratios (HRs) and 95% CIs for CVD incidence per interquartile range (IQR) increment in baseline PFAS concentrations (individual PFAS and PFAS Burden Score), adjusting for a priori confounders (see Covariates section). We validated model assumptions by testing proportional hazards using Schoenfeld residuals and assessing functional forms of continuous covariates using Martingale residuals, with no violations detected. We evaluated potential effect modification by: (1) age (<50, 50–59, and >60 years), (2) sex/menopausal status (male, premenopausal, postmenopausal), (3) treatment arm (placebo, lifestyle intervention, metformin), (4) diet quality (HDI ≤ median versus HDI > median), and (5) physical activity (<2 times per week versus ≥2 times per week) by adding a multiplicative term between plasma PFAS concentrations and the effect modifier. We considered an interaction P value <0.05 to be evidence of effect modification. We also performed stratified analyses by each effect modifier. Due to limited statistical power for individual CVD outcomes, we restricted the stratified analyses to MACE and extended MACE outcomes. Given the multiple statistical tests performed, the potential for type I error inflation was considered. However, as our analyses were primarily hypothesis‐driven, informed by the existing literature on specific PFAS and cardiometabolic risk, we prioritized the interpretation of effect estimates and their CIs over strict adjustment for multiple comparisons. This approach minimizes overcorrection and the potential for type II errors, allowing for a more nuanced evaluation of associations consistent with prior biological knowledge.

We employed quantile g‐computation ³² to evaluate the mixture effect of all 6 PFAS as a mixture. This method estimates the health impact of increasing all PFAS chemicals simultaneously, which better reflects real‐world exposure patterns where people are exposed to multiple chemicals at once. The model works by creating a single “PFAS mixture score” where each person’s exposure to each PFAS is transformed into a rank (eg, low, medium, high) and then summed. It then estimates the change in risk associated with a one‐quantile increase in all PFAS together. Furthermore, the model quantifies the relative contribution (or weight) of each individual PFAS to the overall mixture effect, indicating which chemicals are the most important drivers of the association. We ran models setting the quantile to deciles and incorporated 10 000 bootstraps to ensure robust estimates. We assumed additive effects between PFAS (no interactions) and linear exposure–response relationships. The joint effect is interpreted as mean change in CVD hazard if all plasma PFAS concentrations increased simultaneously by 1 decile.

In sensitivity analyses, we evaluated additional adjustment for education, marital status, physical activity, alcohol consumption, fish intake, and baseline body mass index. We excluded these variables from final models due to minimal impact on effect estimates (<10% change) and to preserve model parsimony. We performed all statistical analyses using R version 4.4.1 (R Foundation for Statistical Computing). This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology guideline. ³³

Results

Study Population

Our analysis included 1382 DPP/DPPOS participants with baseline plasma PFAS concentrations and CVD outcome data (Table 1). The cohort was 65.1% women, and we observed significantly lower PFAS concentrations in women compared with men (tertile 3: 54.5% female versus tertile 1: 79.3% female; P<0.01). Participants in higher PFAS burden tertiles were more likely to be randomized to metformin; to be older, postmenopausal, of Black or non‐Hispanic race, current/former smokers; and to have higher household income.

Median plasma PFAS concentrations (Table 2) were comparable to serum concentrations reported in the general US population of NHANES 1999 to 2000 ³ ; the median PFAS Burden Score was 1.40 (z‐score compared with the NHANES 2017–2018 population mean). Inter‐PFAS correlations varied substantially (Spearman correlation coefficient range: 0.06–0.62; Figure S3), suggesting distinct exposure pathways or sources for different PFAS.

Table 2.

Distributions of Baseline (1996–1999) Plasma PFAS Concentrations (in ng/mL) and PFAS Burden Score Among 1382 Participants From DPP and DPPOS

	Median (Q1, Q3)	Detection frequency, %
PFAS Burden Score	1.40 (1.20, 1.59)	N/A
PFOS	27.9 (18.6, 40.2)	100
PFOA	4.77 (3.47, 6.60)	100
PFHxS	2.40 (1.43, 3.90)	99.9
MeFOSAA	1.10 (0.70, 1.80)	97.9
EtFOSAA*	1.10 (0.60, 2.10)	96.5
PFNA	0.60 (0.40, 0.90)	95.2

DPP indicates Diabetes Prevention Program; DPPOS, Diabetes Prevention Program Outcome Study; MeFOSAA, 2‐(N‐methyl‐perfluorooctane sulfonamido) acetate; PFAS, perfluoroalkyl and polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoate. N/A not applicable; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate; Q1, quartile 1; and Q3, quartile 3.

^*2‐(N‐ethyl‐perfluorooctane sulfonamido) acetate was not measured among the 432 participants in the metformin arm. Limit of detection was 0.1 ng/mL for all PFAS examined.

During a median follow‐up of 21 years, there were 132 MACE (9.6%) and 208 extended MACE (15.1%) events. In crude bivariate analyses without consideration of potential confounders, participants in the highest PFAS burden tertile had more extended MACE outcomes (18.2% in tertile 3 versus 12.3% in tertile 1, P=0.04) and coronary revascularization (8.7% in tertile 3 versus 3.5% in tertile 1, P=0.01).

Prospective Adjusted Associations of PFAS With CVD Incidence

Our analysis showed that higher baseline concentrations of several PFAS were associated with increased cardiovascular risk (Figure 1). Table S1 provides results from unadjusted, age‐adjusted, and fully adjusted models. After adjusting for age, sex, treatment assignment, and lifestyle factors, MeFOSAA showed the strongest association with MACE (HR, 1.16 [95% CI, 1.01–1.33] per IQR increment). In addition, PFOS and EtFOSAA were associated with a 10% to 11% higher risk of extended MACE (PFOS: HR, 1.11 [95% CI, 1.02–1.21]; EtFOSAA: HR, 1.10 [95% CI, 1.03–1.17] per IQR), and MeFOSAA was associated with a 31% increased risk for nonfatal MI (HR, 1.31 [95% CI, 1.12–1.53] per IQR). PFHxS (HR, 1.10 [95% CI, 1.01–1.20]) and EtFOSAA (HR, 1.26 [95% CI, 1.14–1.40] per IQR) were significantly associated with greater risk of congestive heart failure hospitalization. We observed multiple PFAS including PFOA, PFOS, MeFOSAA, and EtFOSAA to be consistently associated with greater risk of CVD death (Figure 1, Table S1). However, we detected no significant associations of PFAS with nonfatal stroke and coronary revascularization (Figure 1, Table S1).

Figure 1. CVD incidence per IQR increment in baseline plasma PFAS concentration among 1382 participants from DPP and DPPOS.

HR (95% CI) estimated using Cox proportional hazard model adjusted for age, sex and menopausal status, race and ethnicity, treatment assignment, income, diet quality, and smoking at baseline. MACE included incidence of nonfatal MI, nonfatal stroke, or CVD death. Extended MACE included MACE and hospitalized CHF, coronary or peripheral revascularization, coronary heart disease by angiography, silent MI. CHF indicates congestive heart failure; CVD, cardiovascular disease; DPP, Diabetes Prevention Program; DPPOS, Diabetes Prevention Program Outcome Study; EtFOSAA, 2‐(N‐ethyl‐perfluorooctane sulfonamido) acetate; HR, hazard ratio; IQR, interquartile range; MACE, major adverse cardiovascular events; MeFOSAA, 2‐(N‐methyl‐perfluorooctane sulfonamido) acetate; MI, myocardial infarction, PFAS, perfluoroalkyl and polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoate; PFOA, perfluorooctanoate; and PFOS, perfluorooctane sulfonate.

Notably, most PFNA estimates trended toward protective associations with multiple cardiovascular outcomes, although none reached statistical significance. Given previous findings in this cohort linking PFNA to fish consumption, ³⁰ we initially hypothesized that dietary factors might explain these null/negative associations. However, additional adjustment for fish intake did not change the effect estimates (Figure S4).

All prospective associations remained robust in sensitivity analyses that accounted for additional covariates, including educational attainment, marital status, body mass index, physical activity, alcohol consumption, and fish intake (Figure S4).

Association of PFAS Mixture With CVD Incidence

Using 2 complementary approaches—quantile‐based g‐computation and PFAS Burden Scores—we found no significant associations between the 6 PFAS mixture and CVD outcomes, with consistently null effects among all models (unadjusted, age‐adjusted, and fully adjusted; Table 3). Given the lack of association of PFNA with any CVD outcomes, we conducted sensitivity analyses excluding PFNA from the mixture specification. These analyses produced virtually identical results (data not shown), suggesting these null findings are robust to mixture specifications.

Table 3.

CVD Incidence Per Quantile of the Baseline PFAS Mixture (Quantile g‐Computation) and Per IQR Increment of PFAS Burden Score Among 1382 Participants From DPP and DPPOS

	Unadjusted	Age‐adjusted	Final covariate–adjusted model
	HR (95% CI)	HR (95% CI)	HR (95% CI)
MACE
PFAS mixture*	1.04 (0.93–1.16)	1.05 (0.94–1.17)	1.00 (0.88–1.13)
PFAS Burden Score†	1.08 (0.93–1.25)	0.99 (0.86–1.15)	0.98 (0.84–1.15)
Extended MACE
PFAS mixture	1.06 (0.97–1.17)	1.07 (0.98–1.17)	1.03 (0.93–1.14)
PFAS Burden Score	1.09 (0.97–1.23)	1.00 (0.89–1.13)	0.97 (0.86–1.10)
Nonfatal MI
PFAS mixture	1.07 (0.90–1.28)	1.08 (0.91–1.29)	1.10 (0.90–1.34)
PFAS Burden Score	1.09 (0.87–1.36)	1.02 (0.82–1.27)	0.99 (0.78–1.26)
Nonfatal Stroke
PFAS mixture	0.94 (0.76–1.17)	0.96 (0.77–1.19)	0.94 (0.74–1.20)
PFAS Burden Score	1.07 (0.79–1.47)	0.98 (0.72–1.32)	1.01 (0.73–1.40)
Hospitalized CHF
PFAS mixture	1.07 (0.88–1.30)	1.09 (0.89–1.32)	1.03 (0.83–1.28)
PFAS Burden Score	0.94 (0.73–1.21)	0.87 (0.69–1.11)	0.90 (0.70–1.16)
Coronary revascularization
PFAS mixture	1.10 (0.95–1.26)	1.11 (0.96–1.29)	1.16 (0.97–1.37)
PFAS Burden Score	1.14 (0.95–1.37)	1.07 (0.89–1.28)	1.00 (0.82–1.22)
CVD death
PFAS mixture	1.09 (0.92–1.28)	1.10 (0.93–1.31)	0.99 (0.81–1.20)
PFAS Burden Score	1.03 (0.82–1.31)	0.94 (0.75–1.18)	1.02 (0.80–1.31)

The age‐adjusted model is adjusted for age only and the final covariate‐adjusted model is adjusted for age, sex and menopausal status, race and ethnicity, treatment assignment, income, diet quality, and smoking at baseline.

CHF indicates congestive heart failure; CVD, cardiovascular disease; DPP, Diabetes Prevention Program; DPPOS, Diabetes Prevention Program Outcome Study; HR, hazard ratio; IQR, interquartile range; MACE, major adverse cardiovascular events; MI, myocardial infarction; and PFAS, perfluoroalkyl and polyfluoroalkyl substances.

^*PFAS mixture includes perfluorooctane sulfonate, perfluorooctanoate, perfluorohexane sulfonate, perfluorononanoate, 2‐(N‐methyl‐perfluorooctane sulfonamido) acetate, 2‐(N‐ethyl‐perfluorooctane sulfonamido) acetate, and PFNA.

^† PFAS Burden Score calculated using the US PFAS exposure burden calculator based on 2017 to 2018 National Health and Nutrition Examination Survey concentrations (https://pfasburden.shinyapps.io/app_pfas_burden/).

Modification of PFAS‐CVD by Demographic Characteristics

Interaction and stratified analyses did not identify any consistent effect modifiers of the PFAS–CVD associations (Tables S2–S6). Age was a borderline significant effect modifier of the MeFOSAA–MACE association (P=0.07 for interaction), with participants older than 60 years at baseline showing the strongest effect (Table S2). Treatment assignment modified the PFOA–MACE association (P=0.05 for interaction) in a direction contrary to our a priori hypothesis, with the strongest association among individuals in the lifestyle intervention arm (P=0.05 for interaction) (Table S3). Diet quality modified the PFHxS–MACE (P=0.03 for interaction) and PFOA–MACE (P= 0.10 for interaction) associations, again contrary to our a priori hypothesis, with the strongest associations among individuals in the higher diet quality group. However, for the PFNA‐extended MACE association, those in the lower diet quality had the strongest association (P=0.09 for interaction) (Table S4). Sex/menopausal status (Table S5) and physical activity (Table S6) did not significantly modify any associations.

Discussion

Our study provides novel evidence that higher plasma concentrations of select PFAS are prospectively associated with greater CVD risk in adults with prediabetes, a metabolically high‐risk population. During a median follow‐up of 21 years, we observed that individual PFAS, particularly MeFOSAA, EtFOSAA, PFHxS, PFOS, and PFOA, were associated with greater risk of incident CVD, including MACE, CVD death, nonfatal MI, and congestive heart failure hospitalization.

These findings align with previous analyses in the same cohort, linking plasma PFAS concentrations to cardiometabolic risk conditions, including dyslipidemia, ¹⁷ diabetes, ³⁴ and coronary artery calcification. ¹⁸ Our results are also consistent with epidemiological studies in other settings. For instance, analysis of NHANES cross‐sectional data (1999–2014, N=10 859) based on self‐reported outcomes found positive associations between serum concentrations of PFOA and PFOS and cardiovascular risk. ¹³ , ¹⁴ , ³⁵ Similarly, an ecological cohort study in Italy’s Veneto region (N=59 147, 1985–2018) ³⁶ reported higher CVD mortality among populations highly exposed to PFAS contamination in water. ³⁶

The observed associations between PFAS and CVD outcomes are biologically plausible. Mechanistic studies indicate PFAS disrupt lipid metabolism by altering peroxisome proliferator‐activated receptor α pathways, leading to elevated low‐density lipoprotein cholesterol and triglycerides, ² , ³⁷ effects consistently reported in DPP/DPPOS¹⁷ and in high‐fat diet rodent models. ³⁸ , ³⁹ PFAS also impair endothelial function by inducing oxidative stress and chronic inflammation, which may accelerate atherosclerosis and plaque instability. ³⁷ In addition, PFAS incorporate into platelet membranes, promoting hypercoagulability and thrombus formation, which may explain the elevated risks of MI that we observed. ²

Our study found no significant associations between the 6 examined PFAS and either nonfatal stroke or coronary revascularization outcomes. While our study population had extensive out‐of‐study use of statins and antihypertensive agents, which could modulate cardiovascular risk, the lack of association between PFAS and stroke aligns with several prior studies. Particularly, the C8 Health Project population in a PFOA‐highly exposed geographic region of the United States found no significant association of PFAS with stroke in participants without diabetes, and inverse associations for PFHxS and PFOS with stroke among participants with diabetes. ⁴⁰ , ⁴¹ , ⁴² Similarly, a Swedish nested case‐control study using data from the Swedish Mammography Cohort and a cohort of 60‐year‐old patients also reported no association of PFAS with stroke (n=1528). ⁴² In contrast, Feng et al found an elevated prevalence of stroke linked to PFOS and PFNA in a cross‐sectional analysis of NHANES 2003 to 2012 data, particularly among men. ⁴³ However, a meta‐analysis incorporating these studies showed no overall association. ⁴⁴ Further investigations would benefit from elucidating mechanistic pathways while prioritizing longitudinal designs to establish temporality and rigorously ascertain outcomes (eg, via adjudicated events or standardized criteria) to minimize survival/reporting bias.

We did not find consistent evidence of effect modification of the PFAS–CVD association by age, sex/menopausal status, treatment arm, diet quality, or physical activity. We found that the PFOA–MACE association was most pronounced in the lifestyle intervention group, although the underlying mechanism remains unclear. We also observed some effect modification by diet, although the direction varied among different PFAS. These findings should be interpreted with caution, as stratification reduced our statistical power to detect modest effects and findings were not consistent among PFAS and outcomes. Further studies with larger sample sizes and longitudinal designs are needed to clarify these potential modifying effects.

Our longitudinal study design, with extended follow‐up (median 21 years) in a high‐risk population with prediabetes and rigorously adjudicated CVD outcomes, has several advantages over previous investigations that were primarily cross‐sectional and used self‐reported CVD outcomes. However, our study also had several limitations. First, single baseline PFAS measurements may not capture long‐term exposure variability, although 4 of the 6 PFAS examined have long half‐lives and our prior work demonstrated stable exposure estimates during 15 years despite declining absolute concentrations, suggesting these measurements could represent long‐term exposures. ³ In addition, we only measured PFAS in a subset of participants who had enough blood sample in the repository, which may limit the generalizability of our findings to the full DPP cohort. The analyzed subgroup differed slightly from the overall cohort in key baseline characteristics, including younger age, higher prevalence of current smoking, and elevated body mass index. Second, residual confounding from unmeasured lifestyle factors cannot be ruled out, such as exposures to other environmental toxicants, neighborhood‐level socioeconomic influences, dietary changes during follow‐up, and physical activity patterns not captured by baseline assessments. Third, our findings may have limited generalizability since all participants had prediabetes and many achieved weight loss through lifestyle intervention. However, the clinical relevance remains substantial given that 1 in 3 US adults have prediabetes ⁴⁵ and weight loss interventions are now more common. Future studies replicating these results in an independent cohort are warranted to strengthen causal inference, although a comparable cohort for this purpose is not presently available.

In conclusion, our study provides longitudinal evidence that specific PFAS can elevate CVD risk in adults with prediabetes. Future studies with larger samples in general population cohorts will additionally help to inform PFAS policies and clinical care.

Sources of Funding

This study was supported by the National Institutes of Environmental Health Sciences (5R01ES024765). Dr Fleisch is supported by a grant from the NIH (R01ES030101).

Disclosures

None.

Disclaimer

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC. Use of trade names is for identification only and does not imply endorsement by the CDC, the Public Health Service, or the US Department of Health and Human Services. The authors declare no competing financial interest.

Supporting information

Tables S1–S6

Figures S1–S4