Association between perfluoroalkyl and polyfluoroalkyl substances and response to the COVID-19 vaccine in a paediatric prospective observational cohort in Arizona

Abstract

Background

Previous research has shown a significant association between increased serum perfluoroalkyl and polyfluoroalkyl substances (PFAS) concentration and reduced antibody response to vaccinations in children, but this association has not yet been studied for COVID-19 vaccinations. We explored this relationship using data from the Arizona Healthcare, Emergency Response and Other Essential Workers Surveillance (AZ-HEROES) Kids study, a prospective cohort that began recruiting participants in July 2021 and followed through April 2023.

Methods

AZ-HEROES Kids participants aged 6 months to 17 years were eligible for this study if they received a full primary mRNA COVID-19 BNT-162b2 (Pfizer-BioNTech) vaccine series, voluntarily submitted a blood specimen 14–60 days following the second vaccine dose and had no evidence of SARS-CoV-2 infection prior to vaccination. 13 PFAS chemicals were measured in serum samples, including branched and linear isomers of perfluorooctanesulfonic acid (PFOS) and perfluorooctanoic acid. A semi-quantitative ELISA was used to measure antibody binding to the SARS-CoV-2 spike protein receptor binding domain (RBD) and S2 subunit domain (S2). An area under the serial dilution curve (AUC) was calculated for RBD and S2. Regression models were fit for each PFAS chemical to assess the relationship between antibody binding (RBD and S2 AUC) and serum PFAS concentration with adjustments for age, sex, geographic region in Arizona and presence of chronic conditions.

Results

The final sample included 120 individuals between 1 and 16 years old. An SD increase in total log-PFOS concentration was significantly associated with a 5.0% decrease in RBD AUC (95% CI −8.1% to −1.8%). The linear and branched isomers of PFOS were also significantly associated with a 4.8% (−7.9% to −1.6%) and 4.9% (−8.0% to −1.7%) decrease in RBD AUC, respectively.

Conclusions

This study contributes to the growing evidence of the negative effects of PFAS exposure on humoral response to vaccination in children.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• In this cohort study of 120 children, increased serum concentration of perfluorooctanesulfonic acid (PFOS) was significantly associated with a dampened immune response following vaccination, as measured via a semi-quantitative ELISA.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Exposure to PFOS may be immunosuppressive for children receiving the COVID-19 vaccine.

Introduction

Perfluoroalkyl and polyfluoroalkyl substances (PFAS) are a group of synthetic chemicals widely used in consumer and industrial products due to their water-resistant, heat-resistant, stick-resistant and stain-resistant properties. PFAS are highly stable because of the presence of carbon-fluorine bonds, giving them robust surfactant properties along with thermal and chemical stability.¹ These substances persist in the environment and can be absorbed through epithelial cells through physical contact or on exposure to contaminated food, water, particle inhalation and dermal exposure.²,⁴ Following absorption, PFAS can bioaccumulate in human tissues and blood causing potential disruption of endocrine systems and other biological functions.⁵,⁷ The National Toxicology Programme has concluded that perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) are presumed to be immune hazards to humans based on substantial evidence from animal studies and a moderate level of evidence from human studies.⁸ Emerging evidence from in vitro, in vivo and epidemiological studies has also raised concerns about the adverse health effects of PFAS exposure, particularly regarding immune function disruption.^{9 10}

There is strong evidence of a relationship between PFAS exposure and reduced immune response to several vaccinations in children. A randomised controlled trial study in Guinea-Bissau found elevated concentrations of PFOS and perfluorodecanoic acid (PFDA) were significantly associated with lower concentrations of post-vaccination measles antibodies.¹¹ Similarly, a Norwegian study found higher concentrations of serum perfluorohexanesulfonic acid (PFHxS), PFOS, PFOA and perfluorononanoic acid (PFNA) reduced antibodies following vaccination to measles.¹² Lastly, a cross-sectional study found increased exposure to several PFAS was correlated with lower antibody levels to mumps and rubella in adolescents.¹³ Overall, while a growing body of evidence supports an inverse association between PFAS exposure and vaccine antibody responses in children, prior studies have reported some inconsistencies across PFAS compounds, vaccines, exposure windows and exposure levels, highlighting the need for further research.¹⁴ In particular, the specific exposure window most relevant to immune outcomes remains uncertain, with studies evaluating prenatal, early childhood and concurrent exposure periods.

To date, most studies evaluating the relationship between PFAS and COVID-19 have focused on adults. A case-control study in China found a relationship between increased urine PFAS concentration and an increased risk of SARS-CoV-2 infection.¹⁵ An ecological study from Sweden also suggested a potential link between high PFAS exposure and susceptibility to SARS-CoV-2.¹⁶ These studies provide some evidence that PFAS exposure may influence the clinical course of COVID-19 by the effect of immunosuppression that diminishes the antibody response to vaccination, particularly with exposure to PFOA, PFOS and PFHxS.¹⁷ However, there is less evidence of a relationship between antibody response to the COVID-19 vaccine and PFAS in adults. A study of individuals in Michigan aged 12–90 years with a history of elevated exposure to PFAS in drinking water found no association between serum PFAS concentrations and antibody response to COVID-19 vaccination.¹⁸ A study in Sweden also found no link between PFAS exposure and lower serum IgG antibody levels against the SARS-CoV-2 spike antigen after two doses of Moderna mRNA vaccine in adults.¹⁹ In Wisconsin, a study found that the relationship between antibody response to vaccination and IQR differences in serum concentration of PFOS, PFOA, PFHxS and PFNA were inversely related, though the relationship was not statistically significant.²⁰ A study on frontline and essential workers in six US states found an association between serum PFOS, PFOA, PFHxS and PFNA concentrations and SARS-CoV-2 antibody levels following infection, but there were no statistically significant relationships between serum PFAS concentration and peak antibody response after vaccination.²¹

Given the limited research on the association of PFAS exposure and immune response to COVID-19 vaccination in children, this study aimed to explore the relationship between PFAS serum levels and antibody response to COVID-19 vaccination in children. Generating robust evidence to examine this relationship would contribute to a growing body of literature aimed at determining the clinical significance of PFAS exposure and bioaccumulation in children. We hypothesised that children with higher levels of serum PFAS concentrations would have lower antibody levels post-vaccination.

Methods

Study design and participant involvement

The AZ-HEROES (The Arizona Healthcare, Emergency Response and Other Essential Workers Surveillance) study is a prospective cohort of adults initiated in 2020 across Arizona that aimed to study the epidemiology and immunology of SARS-CoV-2 infection among individuals with high occupational exposure. In July 2021, the study was expanded to include children aged 6 months to 17 years as part of an extension of the AZ-HEROES study.^{22 23} Recruitment included children of AZ-HEROES participants as well as children from the community.

Data collection

On enrolment, participants completed a survey that included sociodemographic information, SARS-CoV-2 infection history and medical history (including any existing chronic conditions and daily medication use). COVID-19 vaccination status was collected via surveys sent to parents/guardians when the child became eligible to receive the vaccine and on a regular basis thereafter. Vaccination availability for children in the USA expanded throughout the study period, with Emergency Use Authorizations for the BNT-162b2 (Pfizer-BioNTech) issued for ages 16 years or older in December 2020, ages 12–15 in May 2021, ages 5–11 in October 2021 and ages 6 months to 4 years in June 2022.^{24 25} Participants were asked to verify their vaccination by uploading a picture of their vaccination card to the survey or by mailing a physical copy and through the state immunisation registry. 93% of individuals verified their vaccine information.

Participants provided a midturbinate nasal specimen each week and on onset of COVID-19-like symptoms that was tested for SARS-CoV-2 at the Marshfield Clinical Laboratory (Marshfield, Wisconsin) using real-time reverse-transcription PCR. Participants completed monthly follow-up surveys to update information such as mask use, time spent in school and in the community and recent exposures. Instructional models varied following closure of in-person instruction in March 2020, with most schools in Pima and Maricopa Counties resuming in-person instruction by the 2021–2022 academic year.²⁶

All enrolled children had the option to have blood collected at enrollment, after any SARS-CoV-2 infection, after the receipt of any COVID-19 vaccine and at the end of the study. Blood collection was not mandatory and required consent from a parent or guardian. The procedure for blood collection is described elsewhere.²²

Inclusion criteria

Participants were included in this analysis if they received the full primary series of an ancestral monovalent mRNA COVID-19 vaccine, had voluntarily provided a blood specimen between 14 and 60 days after the final dose of the primary series and had no indication of a previous SARS-CoV-2 infection at the time of their blood draw.

We assessed indication of a previous infection through weekly nasal specimens, antibody levels on any previous blood collections prior to vaccination and self-reported data prior to enrolment. Depending on the age of the participant and vaccine manufacturer, a primary series consisted of either two or three doses. Although there were participants in the cohort who received a primary series consisting of three doses, none of these participants submitted a blood specimen between 14 and 60 days after the third dose, resulting in the final analytic set comprising only participants who received two doses for their primary series. Additionally, the analysis was restricted to individuals who received the BNT-162b2 (Pfizer-BioNTech) mRNA COVID-19 vaccine because only one individual received a different vaccine.

Semi-quantitative serological measures

Sera were sent to the University of Arizona Genetics Core laboratory for testing using a locally developed and validated semi-quantitative ELISA to measure antibody binding to the SARS-CoV-2 spike protein receptor binding domain (RBD) and S2 subunit domain (S2), as previously described.²⁷ Briefly, five threefold dilutions of sera were made starting at a 1:60 dilution and ending at 1:4860. From the optical density values at each dilution, an area under the serial dilution curve (AUC) was calculated. This unitless and continuous measurement represented a weighted sum of the optical density value at each dilution. AUC values have been shown to have superior coverage probabilities of serial dilution curves.²⁸ In addition to AUC values, a qualitative antibody result was determined from RBD and S2 optical density values, which was used to detect or confirm prior vaccination or infection in addition to self-reported data.²⁷

Serum PFAS measures

All serum samples were measured for PFAS concentration in ng/mL by the New Jersey Department of Health referencing CDC method # 6304.09.²⁹ A total of 20 PFAS were measured. Linear and branched isomers of PFOS and PFOA were quantified, and the total PFOS and PFOA concentrations were calculated as the sum of the isomers. Only PFAS that had greater than 25% of participant samples with concentration levels above the limit of detection (LOD) were included in the final analysis. 13 PFAS met these criteria and included PFDA, perfluoroheptanesulfonic acid, PFHxS, PFNA, linear and branched isomers of PFOS, linear isomer of PFOA, perfluorobutanesulfonic acid, perfluoroheptanoic acid, perfluoroundecanoic acid, 2-(N-methyl-perfluorooctane sulfonamido) acetic acid (Me-PFOSA-AcOH), total PFOS and total PFOA. There were no samples with a detectable concentration of the branched isomer of PFOA, meaning total PFOA was a linear transformation of its linear isomer in our dataset. To avoid repetitive results, we excluded total PFOA from our analysis. The full list of PFAS measured, the LOD for each PFAS, and the percentage of samples greater than the LOD can be found in Online supplemental eTable 1. The National Academies of Sciences, Engineering, and Medicine (NASEM) has previously published recommendations for medical follow-up and exposure reduction based on the sum of 7 serum PFAS concentrations, all of which are included in this analysis.³⁰ Individuals with a summed concentration between 2 and 20 ng/mL are recommended to reduce their exposure, and individuals above 20 ng/mL are recommended for additional screenings.

Statistical analysis

In our analytic set, serum PFAS concentration and antibody levels were measured on the same blood draw. To assess the relationship between serum PFAS concentration and post-vaccination antibody response, we fit several models. First, for the eight PFAS in which greater than 50% of participant samples had concentrations above the LOD, we fit linear regression models with serum concentration of each PFAS as the primary exposure and antibody response, as measured as RBD or S2 AUC, as the outcome. To account for the possibility of a nonlinear association between serum PFAS concentration and antibody response, we separately fit models specifying PFAS concentration as a continuous variable and a categorical quartile variable. For the continuous models, we imputed any PFAS concentrations that were below the LOD as the LOD divided by the square root of 2.³¹ Additionally, due to the right-skewed distribution of serum PFAS concentrations, we first applied a log-transformation to the concentrations, then normalised to z-scores. To aid in interpretation of differences in AUC levels, we applied a log-transformation to the AUC values. Then, we estimated the percent change in geometric mean AUC levels for each SD change in log-transformed PFAS concentration by (exp(β)−1)×100.

In addition to fitting separate models for each PFAS, we also modelled PFAS concentration as a mixture. We transformed seven of the eight PFAS used in the continuous models to quartiles and fit a quantile g-computation model to estimate the mixture effect, with total PFOS excluded to avoid redundancy with branched and linear PFOS.³² We estimated the expected percent change in geometric mean AUC levels for a simultaneous quartile increase in each PFAS. 95% CIs were estimated using the delta method.

For the four PFAS in which 25%–50% of the participant samples had concentrations above the LOD, we fit linear regression models with PFAS concentration specified as a binary variable (above or below the LOD). Similarly, we log-transformed AUC levels and estimated the percent difference in geometric mean AUC levels for those with PFAS concentration above the LOD compared with those with concentration below the LOD.

In total, we fit 21 models each for RBD and S2: 8 in which PFAS concentration was treated as a continuous variable, 8 in which PFAS concentration was categorised into quartiles, 1 in which PFAS was treated as a mixture and 4 in which PFAS concentration was treated as a binary variable. For each model, we calculated crude and adjusted estimates. We included sex at birth, location (Tucson, Phoenix, other), age and presence of at least one chronic condition (including asthma, other chronic lung disease, cancer, type 2 diabetes, obesity, heart disease, hypertension, immunosuppression, kidney disease, liver disease, neurologic or neuromuscular disease or autoimmune disease) in the adjusted models. These variables were selected a priori based on prior evidence of demographic associations with PFAS exposure and a previous analysis using AZ-HEROES Kids antibody data.³³,³⁵ A directed acyclic graph (DAG) presenting the measured and unmeasured potential confounding variables between serum PFAS concentration and antibody response to the COVID-19 vaccine is available in online supplemental eFigure 1. All analyses were conducted using R Statistical Software (V.4.2.0).³⁶

Reporting guideline

We used the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline to draft this manuscript, and the STROBE reporting checklist when editing.³⁷

Results

Out of 2154 individuals enrolled in AZ-HEROES Kids, 120 were included in the final analytic set, with enrolments into AZ-HEROES Kids occurring from July 2021 to October 2022, and blood draws occurring from August 2021 to November 2022 (figure 1). In the analytic set, 91 (76%) were between 5 and 11 years old at the time of vaccination, 64 (53%) were men, 84 (70%) were from Tucson, Arizona, 84 (70%) were non-Hispanic and white and 105 (88%) had no chronic conditions (table 1). The minimum age at vaccination in the analytic set was 21 months, and the maximum age was 16 years. A comparison between individuals in the analytic set and individuals without a post-vaccination blood draw is available in online supplemental eTable 2.

Figure 1. Flowchart for determining the 120 participants from the AZ-HEROES Kids cohort included in the analytic set from August 2021 to November 2022. AZ-HEROES, Arizona Healthcare, Emergency Response and Other Essential Workers Surveillance.

Table 1. Demographic information for the 120 participants in AZ-HEROES Kids included in the analytic set.

Characteristic	Total (N=120)
Age category, n (%)
<5 years	2 (1.7)
5–11 years	91 (75.8)
12–17 years	27 (22.5)
Site, n (%)
Tucson	84 (70.0)
Phoenix	25 (20.8)
Other	11 (9.2)
Sex, n (%)*
Male	64 (53.3)
Female	55 (45.8)
Race/ethnicity, n (%)*
Non-Hispanic, white	84 (70.0)
Hispanic	24 (20.0)
Other	10 (8.3)
Number of chronic conditions, n (%)†
None	105 (87.5)
One or more	15 (12.5)
Number of daily medications, n (%)*‡
None	94 (78.3)
One or more	24 (20.0)
Serum PFAS concentration, geometric mean (gSD), ng/mL
PFDA	0.050 (2.338)
PFHpS	0.039 (2.872)
PFHxS	0.604 (1.846)
PFNA	0.245 (1.583)
Total PFOS	2.437 (1.647)
Br-PFOS	0.997 (1.631)
L-PFOS	1.419 (1.698)
L-PFOA	1.019 (1.459)

^*Percentages for sex, race/ethnicity and number of daily medications do not add to 100% due to missing data. Less than 5% of participants had missing data for these variables.

^†Chronic conditions include asthma, other chronic lung disease, cancer, type two diabetes, obesity, heart disease, hypertension, immunosuppression, kidney disease, liver disease, neurologic or neuromuscular disease or autoimmune disease).

^‡Participants were asked for the number of daily medications prescribed by a doctor.

AZ-HEROES, Arizona Healthcare, Emergency Response and Other Essential Workers Surveillance; Br-, branched isomer; L-, linear isomer; PFAS, perfluoroalkyl and polyfluoroalkyl substances; PFDA, perfluorodecanoic acid; PFHpS, perfluoroheptanesulfonic acid; PFHxS, perfluorohexanesulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctanesulfonic acid.

Geometric mean concentrations for all PFAS, in which at least 50% of the samples were above the LOD, are summarised in table 1, and sample quartile information is summarised in online supplemental eTable 3, online supplemental eFigure 2, online supplemental eFigure 3. 117 of 120 individuals had a summed concentration of the 7 PFAS referenced in the NASEM recommendations between 2 and 20 ng/mL, and none were above 20 ng/mL. The mean concentration of the seven summed PFAS was 4.96 ng/mL.

Table 2 displays the results from the unadjusted and adjusted models for the 8 PFAS in which at least 50% of the samples were above the LOD and PFAS was treated continuously or as a mixture. A one SD increase in log-transformed total PFOS serum concentration was associated with a 5.0% decrease in RBD AUC antibody levels after vaccination (95% CI −8.1% to −1.8%). This association was significant for both the linear and branched isomers of PFOS, with an estimated mean decrease of 4.8% (95% CI −7.9% to −1.6%) and 4.9% (95% CI −8.0% to −1.7%), respectively. However, there was no statistically significant association between serum PFOS concentration and S2 antibody levels, with an estimated mean difference of S2 antibody levels of −2.8% (95% CI −7.7% to 2.3%) for each SD increase in total PFOS concentration. There were no other statistically significant associations for any of the other PFAS and RBD or S2 AUC. This includes the mixture of PFAS, where a simultaneous quartile increase in all PFAS was non-significantly associated with a mean difference of −4.1% (95% CI −9.0% to 1.0%) and −5.3% (95% CI −12.2% to 2.5%) in RBD and S2 AUC values, respectively.

Table 2. Unadjusted and adjusted estimates of changes in RBD and S2 AUC values for all PFAS in which at least 50% of the samples had a serum concentration above the limit of detection.

PFAS	RBD AUC (95% CI)		S2 AUC (95% CI)
	Unadjusted	Adjusted	Unadjusted	Adjusted
PFDA*	−3.8% (−7.1% to −0.3%)†	−2.1% (−5.3% to 1.3%)	−4.7% (−9.4% to 0.2%)	−3.6% (−8.4% to 1.4%)
PFHpS*	−4.7% (−8.0% to −1.3%)†	−3.3% (−6.7% to 0.3%)	−5.2% (−9.9% to −0.3%)†	−2.1% (−7.4% to 3.5%)
PFHxS*	−1.0% (−4.6% to 2.6%)	−1.8% (−5.3% to 1.8%)	−2.3% (−7.1% to 2.8%)	−0.8% (−6.1% to 4.8%)
PFNA*	−2.3% (−5.8% to 1.3%)	−0.6% (−3.9% to 2.8%)	−1.0% (−6.0% to 4.2%)	0.0% (−4.9% to 5.2%)
Total PFOS*	−4.9% (−8.2% to −1.5%)†	−5.0% (−8.1% to −1.8%)†	−3.8% (−8.6% to 1.2%)	−2.8% (−7.7% to 2.3%)
Br-PFOS*	−4.1% (−7.5% to −0.7%)†	−4.9% (−8.0% to −1.7%)†	−3.5% (−8.3% to 1.5%)	−2.6% (−7.5% to 2.6%)
L-PFOS*	−5.0% (−8.3% to −1.6%)†	−4.8% (−7.9% to −1.6%)†	−3.6% (−8.3% to 1.4%)	−2.6% (−7.5% to 2.5%)
L-PFOA*	−4.7% (−8.0% to −1.3%)†	−1.5% (−5.0% to 2.1%)	−5.1% (−9.7% to −0.2%)†	−1.8% (−6.9% to 3.7%)
Mixture‡	−4.0% (−8.9% to 1.1%)	−4.1% (−9.0% to 1.0%)	−7.4% (−14.0% to −0.4%)†	−5.3% (−12.2% to 2.5%)

^*Serum PFAS concentrations were treated as a continuous variable in the models and were log-transformed and transformed to z-scores. RBD and S2 AUC values were log-transformed. Percentages represent the estimated geometric mean percent change in RBD or S2 AUC values for each SD increase in log-serum PFAS concentration. These estimates were calculated as (exp(β)−1)×100), where β represents the coefficient for PFAS concentration estimated from the linear regression model.

^†Represents statistically significant change in mean AUC values for alpha=0.05.

^‡Serum PFAS concentrations were transformed to quartiles prior to fitting a quantile g-computation model to estimate the effect of the mixture of all PFAS (except total PFOS). Percentages represent the estimated geometric mean percent change in RBD or S2 AUC values for a simultaneous quartile increase in all PFAS. 95% CIs estimated via the delta method.

AUC, area under the serial dilution curve; Br-, branched isomer; L-, linear isomer; PFDA, perfluorodecanoic acid; PFHpS, perfluoroheptanesulfonic acid; PFHxS, perfluorohexanesulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctanesulfonic acid; RBD, receptor binding domain.

Online supplemental eTable 4 displays the results from the analysis in which serum PFAS concentration was categorised into quartiles in the linear regression models. Similarly to the primary analysis, increased serum PFOS concentrations was associated with decreased RBD AUC values after vaccination. In particular, individuals in the third quartile of serum total PFOS concentration had an estimated 9.1% mean decrease in RBD AUC values (95% CI −17.2% to −0.1%) compared with individuals in the lowest quartile, and individuals in the highest quartile had an estimated 10.7% mean decrease in RBD AUC values (95% CI −18.7% to −1.8%) compared with individuals in the lowest quartile. Similar associations were found for the linear and branched isomers of PFOS. Like the primary analysis, there were no other statistically significant associations for any of the other PFAS and RBD or S2 AUC

Table 3 displays the results from the unadjusted and adjusted linear regression models for the four PFAS in which between 25% and 50% of the samples were above the LOD. Although there were no statistically significant associations, those with serum PFUnA concentration above the LOD had an estimated mean decrease of 9.1% in S2 AUC compared with those below the LOD (95% CI −18.4% to 1.1%). Those with serum Me-PFOSA-AcOH concentration above the LOD had an estimated mean decrease of 6.0% in RBD AUC (95% CI −12.5% to 1.1%) and an estimated mean decrease of 8.5% in S2 AUC (95% CI −18.0% to 2.0%) compared with those below the LOD.

Table 3. Unadjusted and adjusted estimates of changes in RBD and S2 AUC values for serum PFAS concentration above the limit of detection versus below the limit of detection for all PFAS in which 25%–50% of the samples had a serum concentration above the limit of detection.

PFAS	RBD AUC (95% CI)		S2 AUC (95% CI)
	Unadjusted	Adjusted	Unadjusted	Adjusted
PFBS	−3.0% (−10.7% to 5.4%)	−0.4% (−7.7% to 7.4%)	0.0% (−11.0% to 12.3%)	0.7% (−10.2% to 12.9%)
PFHpA	2.3% (−5.2% to 10.5%)	5.1% (−2.0% to 12.8%)	1.6% (−8.8% to 13.2%)	4.2% (−6.4% to 15.9%)
PFUnA	−4.3% (−11.7% to 3.6%)	−3.0% (−9.7% to 4.2%)	−10.5% (−19.9% to 0.1%)	−9.1% (−18.4% to 1.1%)
Me-PFOSA- AcOH	−3.9% (−11.3% to 4.1%)	−6.0% (−12.5% to 1.1%)	−7.5% (−17.3% to 3.5%)	−8.5% (−18.0% to 2.0%)

Serum PFAS concentrations were treated as a binary variable in the models (above or below the LOD). LOD values can be found in the online supplemental material. RBD and S2 AUC values were log-transformed. Percentages represent the estimated mean geometric percent change in RBD or S2 AUC values for serum PFAS concentration above the LOD versus below the LOD. These estimates were calculated as [(exp(β)−1])×100]), where β represents the coefficient for the binary PFAS concentration estimated from the linear regression model.

Adjusted linear regression models included sex at birth, location (Tucson, Phoenix, other), age (continuous, years), and presence of at least one chronic condition (binary yes/no).

AUC, area under the serial dilution curve; LOD, limit of detection; Me-PFOSA-AcOH, 2-(N-Methyl-perfluorooctane sulfonamido) acetic acid; PFAS, perfluoroalkyl and polyfluoroalkyl substances; PFBS, perfluorobutanesulfonic acid; PFHpA, perfluoroheptanoic acid; PFUnA, perfluoroundecanoic acid; RBD, receptor binding domain.

Discussion

In this analysis, we found a significant association between increased serum PFOS concentration and a reduced antibody response to the primary series of a COVID-19 mRNA vaccine in a cohort of children aged 1–16 years in Arizona, USA. We observed this relationship when considering RBD AUC values as continuous or categorical, measured via a validated semi-quantitative ELISA as our outcome measure. No other PFAS considered in this analysis was found to be significantly associated with antibody response to vaccination, and there was no significant association between the mixture of PFAS concentrations and antibody response. Additionally, there was no significant relationship between serum PFOS concentration and antibody response quantified by S2 AUC values.

We have previously studied the relationship between antibody response to vaccination among children who later became infected. In a set of 79 children aged 5–11 from the AZ-HEROES Kids cohort and the broader PROTECT cohort, significantly higher RBD AUC values were observed after vaccination among individuals who did not experience a post-vaccination infection compared with those who did.^{22 33} Each SD increase in RBD AUC resulted in an estimated 47% reduction in the odds of post-vaccination infection (aOR 0.53; 95% CI 0.29 to 0.97). RBD AUC values used in this previous analysis came from the same assay used in the present analysis. Our findings, along with these previous findings, suggest that the dampened immune response after vaccination that is associated with increased serum PFOS concentration in children may increase one’s risk for future infection.

The inverse relationship between serum PFOS concentration and antibody response to vaccination observed in this study aligns with previous epidemiological studies assessing PFOS exposure and response to vaccination in paediatric populations.³⁸,⁴⁰ In addition to the evidence of being immunosuppressive in children, PFOS has also been shown to induce epigenetic alterations and oxidative stress.^{41 42} This evidence is consistent with findings from experimental systems.

This analysis is subject to limitations. First, it was limited to children who voluntarily collected a blood specimen following vaccination, and most individuals in AZ-HEROES Kids opted to not collect a blood specimen during this window. Second, most individuals in the analytic set were older than 4 years of age, and therefore results may not be generalisable to younger children. Third, although AZ-HEROES Kids is a prospective cohort, we were limited to cross-sectional PFAS and antibody data for this analysis, which limits the causal interpretation of our results. In particular, a single serum PFAS measurement does not capture temporal variability in exposure or potential developmental windows of susceptibility. However, due to their long biological half-lives, serum PFAS concentrations are widely considered stable biomarkers of cumulative long-term exposure in epidemiologic studies. Fourth, the observed PFAS concentrations in this cohort were generally lower than the concentrations observed in the 2017/2018 National Health and Nutrition Examination Survey (NHANES) for individuals aged 12–19, which is a representative sample for the USA.⁴³ Our results may not be applicable for individuals with a higher burden of PFAS exposure. Fifth, we were limited to measuring 20 PFAS. Although these PFAS included those commonly found in humans, they only consist of a small subset of all PFAS used in products.^{44 45} Sixth, there was potential for misclassification bias if an individual had an infection prior to vaccination that was not detected. Finally, all hypothesis testing conducted in this analysis is subject to a potentially inflated type I error due to multiple comparisons.

This analysis has several strengths. First, although we relied on voluntary specimen collections, we successfully generated a large sample size of 120 individuals, which is particularly valuable for studies analysing paediatric serum PFAS concentrations as sample size is often a primary limitation. Second, there were only minimal missing data in the variables used in the adjusted models (<5%), reducing the potential bias and improving the robustness of the analysis outcome. Third, we were able to closely monitor participants’ infection status through weekly swab collections, enabling us to confidently identify and select individuals with no prior infections in the analytic set. Finally, blood draws were collected 14–60 days post vaccination, allowing for controlled comparisons between groups.

This study contributes to the growing literature assessing PFAS exposure and blunted antibody response after vaccination in children. To our knowledge, this is the first study to assess the association between serum PFAS concentration and antibody response following COVID-19 vaccination in a paediatric cohort. While we observed a significant association between PFOS and a lower antibody response following vaccination in our cohort, longitudinal studies with repeated measurements are needed to better understand the relationship of PFAS and immune response and to adjust for time-varying confounders. Additionally, future research should evaluate the clinical significance of the association found in this analysis, including the relationship between differences in antibody response associated with increased PFAS exposure and incidence of SARS-CoV-2 infection. Such work can help clarify the complicated relationship between PFAS exposure, antibody response and clinical outcomes.

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the US Centers for Disease Control and Prevention.

Data availability statement

Data may be obtained from a third party and are not publicly available.