Comparison of Serum PFAS Concentrations in Incumbent and Recruit Firefighters and Longitudinal Assessment in Recruits

Abstract

Objective:

Firefighters are occupationally exposed to per- and polyfluoroalkyl substances (PFAS). This study objective was to compare serum PFAS concentrations in incumbent and recruit firefighters and evaluate temporal trends among recruits.

Methods:

Serum PFAS concentrations were measured in 99 incumbent and 55 recruit firefighters at enrollment in 2015-2016, with follow-up 20-37 months later for recruits. Linear and logistic regression and linear mixed-effects models were used for analyses. Fireground exposure impact on PFAS concentrations was investigated using adjusted linear and logistic regression models.

Results:

Incumbents had lower n-PFOA and PFNA than recruits and most PFAS significantly decreased over time among male recruits. No significant links were found between cumulative fireground exposures and PFAS concentrations.

Conclusions:

Serum PFAS concentrations were not increased in incumbent firefighters compared with recruits and were not associated with cumulative fireground exposures.

Introduction

Per- and polyfluoroalkyl substances (PFAS) are a broad group of thousands of chemicals consisting of carbon-alkyl chains of variable length in which some or all of the hydrogen atoms have been replaced by fluorine atoms.¹ Substituting fluorine for hydrogen atoms produces highly stable molecules, with high thermal and chemical resistant properties.² The attachment of a charged moiety (e.g., carboxylate or sulfonate) produces a soluble molecule that has both hydrophobic and lipophobic properties, allowing for these chemicals to persist in the environment and, for some, to bioaccumulate in various ecosystems.^1,3

PFAS have been widely used in the production of industrial and consumer products for over sixty years. Ubiquitous detection of PFAS in the environment and in humans has raised concerns about the long-term effects of these chemicals on human health. Long-alkyl chain PFAS have carbon-chain lengths ≥ 7 for perfluorocarboxylic acids and ≥ 6 for perfluorosulfonic acids. They have been largely phased out of production and are often referred to as legacy PFAS. Serum concentrations of legacy PFAS such as perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS) are generally declining in the general United States (US) population,⁴ but less is known about changes over time in firefighters given ongoing occupational exposures. Although the production of long-alkyl chain PFAS has decreased as a result of regulatory efforts in several countries including the US,¹ manufacturing persists in some countries, and the US continues to produce novel alternatives to long-alkyl chain PFAS.³ PFAS are still being used to make substances such as home goods, medical devices, treated textiles, electronics, printing products, surface treatments, and certain Class B fire-suppression foams (e.g. aqueous film-forming foam, or AFFF).^5,6 Legacy PFAS-containing AFFF was produced and sold in the US up until 2016; and although long-alkyl chain PFAS are currently not being manufactured in the US, precursor compounds contained in modern fluorotelomer AFFF may undergo transformations to form legacy PFAS.^1,7-9 In addition, the use of stockpiled legacy PFAS-containing AFFF has continued to be used for emergency fire suppression purposes in some jurisdictions.¹⁰ Currently, manufacturers of modern PFAS-containing products are not required to disclose specific PFAS, with the exception of PFOS salts and their precursors.⁷ A study that performed untargeted analysis reported PFAS-like features in several AFFF foams.^7,11 However, because specific chemical compositions are based on proprietary blends, there is limited information as to which specific fluorosurfactants are currently being used in modern AFFF formulations.

Exposure to many legacy PFAS can lead to accumulation over time in animal and human tissues.¹² Detectable serum concentrations of PFOA, PFOS, perfluorononanoate (PFNA), and perfluorohexane sulfonate (PFHxS) have been measured in humans worldwide.¹³ Elevated PFAS concentrations in humans have been associated with adverse health effects including but not limited to elevated cholesterol; cardiovascular, kidney and thyroid diseases; cancers of the prostate, kidney, testes, and breast; and decreased immune response.^14-23

Firefighting is a uniquely hazardous occupation with various known exposures and health risks, and firefighters are more likely than the public to develop multiple cancers, including urogenital cancers of the testis, prostate and kidney,^24,25 cancers which are associated with PFAS exposure in other populations.^15-18,26,27 Firefighters may be particularly at risk of higher serum PFAS concentrations based on their potential for direct exposure to AFFF,^28,29 equipment and turnout gear containing PFAS,³⁰ fire station dust containing PFAS,³¹ and combustion products produced from burning PFAS-coated materials.³² The extent of exposure to PFAS may depend on the number of fire incidents they experience, as well as the duration and type of firefighting events.³³ Incumbent firefighters (e.g., career and supervisory firefighters) may have greater cumulative exposures to occupational PFAS sources, compared to recruit firefighters who are often new to the fire service.

Given the potential for increased PFAS exposure in firefighters, the objectives of this study were to compare serum PFAS concentrations in incumbent and recruit (prior to live-fire training) firefighters at study enrollment. We also sought to evaluate overall changes in serum PFAS over time in the recruit firefighters and the relationships between serum PFAS and both cumulative and most recent fireground exposure. Understanding the sources and patterns of PFAS exposure among firefighters can inform policies aimed at reducing exposure and highlight areas for targeted intervention.

Methods

Study Population

The study participants were recruited from the Tucson Fire Department in Tucson, Arizona, as part of a comprehensive cancer prevention study among firefighters, as previously described.^34-38 The study included both incumbent and recruit firefighters. The recruitment of recruits took place before their occupational live-fire training. All participants completed an online exposure assessment survey, which collected basic demographic information (such as date of birth, sex, race, and ethnicity) and health-related data (such as weight and height for body mass index calculation). Biological samples were collected from all participants during enrollment between 2015 and 2016. For recruit firefighters, the survey and biological samples were repeated once during the follow-up period of the study between 2017 and 2018. Race, ethnicity, and sex were self-reported by the participants. All participants provided informed consent in accordance with the protocols approved by the University of Arizona Institutional Review Board (#1509137073).

Fireground Exposure Assessment

Fireground exposure data for recruit firefighters during the follow-up period were collected from fire departmental records. These records contained detailed information on various aspects of firefighting events, such as the count and duration of fire runs, type of fire, and date of most recent fire prior to follow-up blood collection. In this study, we focused on specific fireground exposure variables, namely the total number of fire runs, the total duration (measured in hours) of fire runs, the number of days since most recent fire, the number of structural fire runs, the duration (measured in hours) of structural fire runs, and the number of days since most recent structural fire.

Blood Sample Collection and Quantification of Serum PFAS

Blood samples were collected using 10 mL glass serum collection tubes (Becton, Dickinson and Company, Franklin Lakes, NJ). Following Arizona Department of Transportation guidelines, all blood samples were transferred at 4°C to the University of Arizona. Upon receipt, the whole was centrifuged at 1300xg for 15 minutes, and the resulting serum was aliquoted into 0.5 mL cryogenic tubes. All samples were stored at −80°C until analysis.

The quantification of nine PFAS (PFHxS, linear PFOS [n-PFOS], monomethyl branched isomers of PFOS [Sm-PFOS], linear PFOA [n-PFOA], branched isomers of PFOA [Sb-PFOA], PFNA, perfluorodecanoate [PFDA], perfluoroundecanoate [PFUnDA], and 2-(N-methyl-perfluorooctane sulfonamido) acetate [MeFOSAA]) in blood serum samples was conducted by shipping the samples overnight to the US Centers for Disease Control and Prevention (CDC, Atlanta, Georgia, USA). The quantification procedures followed the methods described by Kato et al., 2018.³⁹ The limit of detection (LOD) for all PFAS in this study was 0.1 nanograms per milliliter (ng/mL). Supplementary Table 1 provides additional characteristics of these nine PFAS.

The analysis of de-identified specimens at the CDC laboratory was determined not to constitute engagement in human subjects’ research.

Statistical Analyses

All statistical procedures were performed using R versions 3.6.0, 3.6.3, 4.0.3, and 4.0.4 (R Foundation for Statistical Computing, Vienna, Austria).⁴⁰ We examined participant demographics and key predictor variables (years as a firefighter and fireground exposure variables for recruits only). These were stratified by biological sex, which was self-reported as either “male” or “female”, and by firefighting status, categorized as “incumbent” or “recruit”. We employed this stratification to provide a nuanced interpretation of the data since sex-based biological differences may influence PFAS metabolism,^41-43 and distinct occupational roles, exposure histories, and length of service could lead incumbent and recruit firefighters to experience different PFAS levels.⁴⁴ Age was calculated as years from date of birth to date of study enrollment. We tested whether demographic characteristics were different by firefighter status (incumbent vs recruit) using Fisher’s exact test for categorical variables and t-tests for continuous variables.

Serum concentrations of PFAS detected in at least 50% of samples were log-transformed and analyzed as continuous variables. For samples below the LOD, a value of 0.07 ng/mL (i.e., LOD/√2) was imputed before log-transformation.⁴⁵ Serum concentrations of PFAS that were non-detectable in more than 50% of samples were dichotomized as detectable or undetectable. We calculated the geometric means (GMs) and geometric standard deviations (GSDs) for continuous PFAS concentrations, and medians with interquartile ranges for dichotomized PFAS, again stratified by biological sex and firefighting status. To examine the correlations among PFAS, we used Pearson’s correlation coefficients for pairs of continuous (log-transformed) PFAS, biserial correlations for continuous-dichotomous PFAS pairs,^46,47 and categorical analyses (log odds ratio) for pairs of dichotomized PFAS. For all demographic and PFAS summaries, information on recruit women is not shown because of low numbers in this stratum and consequent concerns about identifiability.

Cross-Sectional Comparisons of PFAS by Firefighter Status

The concentration data for each PFAS were fit to a series of either linear or logistic regression models with age, sex, and ethnicity (non-Hispanic and Hispanic) selected a priori as covariates based on previously suggested determinants of various PFAS concentrations,^50,51 and firefighter status (incumbent and recruit) as the predictor of interest. Given the potential sex-related differences in PFAS concentrations, sensitivity analyses were also conducted exclusively on male incumbents and male recruits (adjusting for age and ethnicity) to examine whether the male stratum-specific findings are consistent with the overall findings. Four participants (3%) with missing or unknown ethnicity were excluded from analyses. However, in the interest of maintaining data integrity and study transparency, the excluded participants were still included in the data summary tables. Inference for linear regression models was conducted using t-tests. For the logistic regression models, inference was conducted using likelihood ratio tests. We calculated the percent difference between incumbents and recruits using model regression coefficients where percent difference = (exp(β) − 1) × 100.

Longitudinal Assessment of Changes in PFAS Concentration Over Time

Longitudinal analyses were conducted on enrollment and follow-up data from recruits only. These models were exclusively fit to data from male participants due to the limited number of female recruits (n = 3). We used the R package lme4 to fit a linear mixed effects model to each PFAS concentration that was detectable for 50% or more of individuals.⁵² Each model included subject-specific random intercepts (to account for the likely correlation between data from the same participant) and a priori-selected fixed effects for ethnicity (non-Hispanic and Hispanic), years of age at enrollment, and the key predictor variable of time (years) since enrollment. For PFAS with detectable concentrations in less than 50% of individuals (Sb-PFOA and MeFOSAA), we fit logistic regression models using generalized estimating equations (GEE) models with an exchangeable correlation structure to account for intra-subject correlation of repeated PFAS measurements.⁵³ The GEE models included ethnicity (non-Hispanic and Hispanic), sex, and time (years) since enrollment as the key predictor. Inference for the linear mixed effects models was conducted using t-tests with both degrees of freedom and the variance-covariance matrix of fixed effects adjusted via the method of Kenward and Roger using the R package pbkrtest.^54,55 For the GEE models, inference was conducted using Wald χ² tests with robust standard errors.

Associations of Fireground Exposure with PFAS at Follow-Up

To assess the associations between fireground exposure and serum PFAS concentrations at follow-up, we considered six variables as proxies for exposure: overall number of fire runs since enrollment, number of structural fire runs since enrollment, days since most recent fire, days since most recent structural fire, cumulative duration of all fire runs since enrollment, and cumulative duration of structural fire runs since enrollment. Training fire hours were excluded from our exposure metrics.

We fit a series of minimally adjusted and fully adjusted models to each recruit follow-up PFAS concentration. Each of the six minimally adjusted models had one of the six exposure proxies mentioned above as the key predictor. We adjusted for both years since enrollment and the same log-transformed PFAS chemical concentration (for continuous outcomes) or detectability (for dichotomous outcomes) at enrollment. The six corresponding fully adjusted models additionally adjusted for ethnicity (non-Hispanic and Hispanic) and age at enrollment. We used linear regression models for continuous PFAS concentration outcomes and logistic regression models for dichotomous PFAS concentration outcomes. Due to the small number of female recruit participants, all exposure models were exclusively fit to data from recruit males. Following suggestions by Rothman (1990) and Rubin (2021)^48,49, we did not adjust inferential results for multiple comparisons.

Results

Study Participants

The study consisted of 154 participants, and Table 1 presents a summary of their characteristics at enrollment. Additionally, it provides an overview of fireground exposure variables specifically for recruit males during the follow-up period. Most of the participants were male (82%), with approximately 23% identifying as Hispanic. Three female incumbent firefighters and one male incumbent firefighter had unknown or missing ethnicity and were included in data summaries but excluded from regression analyses. Consequently, a maximum of 150 participants were included in the regression analyses. Compared to recruits, incumbent firefighters had a higher mean BMI and were older on average. The average firefighting experience among incumbent firefighters was 14.7 years. Recruits had an average of 26 months to follow-up. During this period, they participated in an average of 48.8 fire runs and accumulated approximately 26.8 hours on the fireground. Structural fires accounted for approximately one-third of their fire runs and approximately half of their fire hours.

Table 1.

Characteristics of incumbent (n = 99) and male recruit firefighters (n = 52) at enrollment and summary of fireground exposures for male recruits at follow-up (n = 52).

Characteristics	Incumbents			Recruits	p-value
	Females (n = 25)	Males (n = 74)	Overall (n = 99)	Males (n = 52)
Years of Age (mean (SD))	44.32 (8.74)	40.59 (8.46)	41.54 (8.64)	28.35 (6.05)	< .01
Sex (N (%))
Female	25 (100.0)	0 (0.0)	25 (25.3)	0 (0.0)	< .01
Male	0 (0.0)	74 (100.0)	74 (74.7)	52 (100.0)
Ethnicity (N (%))
Non-Hispanic	20 (80.0)	53 (71.6)	73 (73.7)	40 (76.9)	0.85
Hispanic	2 (8.0)	20 (27.0)	22 (22.2)	12 (23.1)	0.84
Unsure/Missing	3 (12.0)	1 (1.4)	4 (4.0)	0 (0.0)	0.3
Race (N (%))
White	21 (95.5)	62 (83.8)	83 (86.5)	43 (84.3)	0.64
American Indian/Alaska Native	0 (0.0)	1 (1.4)	1 (1.0)	3 (5.8)	0.06
Black/African American	0 (0.0)	0 (0.0)	0 (0.0)	4 (7.7)	0.02
Other Race	1 (4.5)	14 (18.9)	15 (15.6)	9 (17.3)	0.82
BMI (mean (SD))	26.10 (3.59)	27.62 (3.34)	27.23 (3.45)	26.12 (2.74)	< .01
Years as Firefighter (mean (SD))	16.85 (6.90)	14.01 (7.03)	14.59 (7.05)	0.76 (1.41)	< .01
Months to Follow-Up (mean (SD))				26.09 (4.27)
Number of Fire Runs (mean (SD))				48.81 (15.13)
Hours of Fire Runs (mean (SD))				26.80 (8.76)
Days Since Most Recent Fire Run (mean (SD))				59.28 (63.19)
Number of Structural Fire Runs (mean (SD))				16.76 (6.71)

Note: This table provides an overview of the demographics of incumbent and recruit firefighters at enrollment, including sex-stratified data for incumbent firefighters. It also provides a summary of fireground exposure variables for recruits at follow-up. All values represent data at enrollment. All values represent values at enrollment, except for fireground exposure variables and months to follow-up. To ensure participant confidentiality, demographic characteristics for the three female recruits were excluded due to small numbers. Races with lower representation, such as American Indian or Alaska Native, Asian, and Native Hawaiian or Other Pacific Islander, were combined under Other Race. Body Mass Index (BMI) units are kg/m². Fisher’s exact tests were performed to compare categorical variables between overall incumbents and male recruits at enrollment, and t-tests were used for comparisons of continuous variables. A p-value < .05 was considered statistically significant, indicating a significant difference between the two groups.

Serum PFAS Concentrations

Table 2 presents the serum PFAS concentrations stratified by firefighter status and sex. PFAS compounds that were detected in at least 50% of samples were PFHxS, n-PFOS, Sm-PFOS, n-PFOA, PFNA. PFDA, and PFUnDA. However, PFUnDA exhibited low variability and deviation from the expected normal distribution, necessitating dichotomization. Consequently, PFUnDA was treated as a binary variable, alongside Sb-PFOA and MeFOSAA. At enrollment, Sb-PFOA was detected in 39% of incumbent firefighters and 27% of recruit firefighters, while MeFOSAA was detected in 41% of incumbent firefighters and 55% of recruit firefighters. At follow-up, Sb-PFOA and MeFOSAA were detected in 48% and 38% of male recruit firefighters, respectively. PFHxS, n-PFOS, Sm-PFOS, and n-PFOA had the highest detection frequencies across all groups. Among incumbent firefighters, males had numerically higher GM concentrations than females for PFHxS, n-PFOS, Sm-PFOS, n-PFOA, PFNA, and PFDA. Among recruits, the numerically highest GM concentrations at enrollment (all recruits) and follow-up (males only) were for n-PFOS (4.01 and 3.30 ng/mL, respectively), while the lowest were for Sb-PFOA at enrollment (all recruits) and both Sb-PFOA and MeFOSAA (males only) at follow-up (median less than the LOD for all).

Table 2.

Serum concentrations (ng/mL) of nine PFAS in incumbent and recruit firefighters (N = 154).

Serum PFAS Concentrations		Enrollment				Follow-Up
		Female Incumbents (n = 25)	Male Incumbents (n = 74)	Overall Incumbents (n = 99)	Overall Recruits (n = 55)	Male Recruits (n = 52)
PFHxS	Above LOD (N (%))	25 (100)	74 (100)	99 (100)	55 (100)	52 (100)
	GM (GSD)	1.42 (2.03)	3.22 (1.74)	2.62 (2.00)	2.83 (1.81)	2.85 (1.77)
	Min, Max	0.20, 6.90	0.70, 12.50	0.20, 12.50	0.80, 16.50	0.80, 14.00
n-PFOS	Above LOD (N (%))	25 (100)	74 (100)	99 (100)	55 (100)	52 (100)
	GM (GSD)	2.36 (1.76)	4.20 (1.44)	3.63 (1.64)	4.01 (1.56)	3.30 (1.56)
	Min, Max	0.70, 6.30	1.70, 12.40	0.70, 12.40	2.10, 18.30	1.50, 15.30
Sm-PFOS	Above LOD (N (%))	25 (100)	74 (100)	99 (100)	55 (100)	52 (100)
	GM (GSD)	1.17 (1.88)	2.41 (1.56)	2.01 (1.80)	2.03 (1.53)	1.71 (1.52)
	Min, Max	0.30, 4.00	0.70, 6.60	0.30, 6.60	0.80, 6.50	0.80, 5.40
n-PFOA	Above LOD (N (%))	25 (100)	74 (100)	99 (100)	55 (100)	52 (100)
	GM (GSD)	1.35 (1.83)	1.77 (1.44)	1.65 (1.57)	2.24 (1.49)	1.77 (1.46)
	Min, Max	0.40, 3.90	0.80, 4.80	0.40, 4.80	1.00, 8.60	0.80, 5.20
PFNA	Above LOD (N (%))	21 (84)	74 (100)	95 (96)	55 (100)	52 (100)
	GM (GSD)	0.32 (2.19)	0.42 (1.45)	0.39 (1.68)	0.59 (1.75)	0.39 (1.57)
	Min, Max	< LOD, 1.20	0.20, 1.70	< LOD, 1.70	0.20, 11.00	0.20, 2.30
PFDA	Above LOD (N (%))	25 (100)	74 (100)	99 (100)	55 (100)	52 (100)
	GM (GSD)	0.21 (1.75)	0.25 (1.30)	0.24 (1.44)	0.26 (1.48)	0.22 (1.40)
	Min, Max	0.10, 1.10	0.20, 0.60	0.10, 1.10	0.10, 0.80	0.10, 0.50
PFUnDA	Above LOD (N (%))	17 (68)	46 (62)	63 (64)	32 (58)	28 (54)
	Median (IQR)	0.10 (< LOD, 0.20)	0.10 (< LOD, 0.20)	0.10 (< LOD, 0.20)	0.10 (< LOD, 0.15)	0.10 (< LOD, 0.10)
	Min, Max	< LOD, 0.30	< LOD, 0.70	< LOD, 0.70	< LOD, 0.30	< LOD, 0.30
MeFOSAA	Above LOD (N (%))	10 (40)	31 (42)	41 (41)	30 (55)	20 (38)
	Median (IQR)	0.10 (< LOD , 0.10)	< LOD (< LOD, 0.10)	< LOD (< LOD, 0.10)	0.10 (< LOD, 0.20)	< LOD (< LOD, 0.10)
	Min, Max	< LOD, 2.20	< LOD, 0.90	< LOD, 2.20	< LOD, 0.90	< LOD, 0.80
Sb-PFOA	Above LOD (N (%))	13 (52)	26 (35)	39 (39)	15 (27)	25 (48)
	Median (IQR)	< LOD (< LOD , 0.10)	< LOD (< LOD, 0.10)	< LOD (< LOD, 0.10)	< LOD (< LOD, 0.10)	< LOD (< LOD, 0.10)
	Min, Max	< LOD, 0.20	< LOD, 1.70	< LOD, 1.70	< LOD, 0.60	< LOD, 0.50

Note: This table provides a summary of the serum concentrations of per- and polyfluoroalkyl substances (PFAS) measured in incumbent and recruit firefighters at enrollment, as well as in recruit males at follow-up. Sex-stratified results are provided for incumbent firefighters, while sex-stratified results for recruits were not reported due to the small number of female recruits (n = 3). PFAS serum concentrations are reported in nanograms per milliliter (ng/mL). Central tendency and variability are represented using the geometric mean (GM) and geometric standard deviation (GSD) for PFAS detected in 50% or more of samples, except for PFUnDA, which exhibited low variability. Central tendency and variability for MeFOSAA and Sb-PFOA, detected in less than 50% of samples, along with PFUnDA, are represented using the median and interquartile range (IQR). The number and percentage of samples detected above the limit of detection (LOD) are presented for each compound, as well as the minimum and maximum measured concentrations. The LOD for all PFAS compounds was 0.1 ng/mL The measured PFAS compounds include perfluorohexane sulfonate (PFHxS), linear perfluorooctane sulfonate (n-PFOS), perfluoromethylheptane sulfonate isomers (Sm-PFOS), linear perfluorooctanoate (n-PFOA), branched PFOA isomers (Sb-PFOA), perfluorononanoate (PFNA), perfluoroundecanoate (PFUnDA), and 2-(N-methyl-perfluorooctane sulfonamide) acetate (MeFOSAA).

Correlations among PFAS are presented in Supplementary Information Figure 1. Significant positive (p < 0.05) correlations were observed between log-transformed serum concentrations of PFHxS, n-PFOS, Sm-PFOS, n-PFOA, PFNA, or PFDA. Biserial correlation analysis between detectability of MeFOSAA, Sb-PFOA, or PFUnDA and continuous (log-transformed) PFAS revealed significant negative correlations between Sb-PFOA and n-PFOS or PFDA (p < 0.05), positive correlations between MeFOSAA and n-PFOS, n-PFOA, or PFNA (p < 0.05), and significant positive correlations between PFUnDA and PFHxS, n-PFOS, Sm-PFOS, PFNA, or PFDA (p < 0.05). There were no significant correlations between detectable and non-detectable concentrations of Sb-PFOA and MeFOSAA, Sb-PFOA and PFUnDA, or MeFOSAA and PFUnDA.

Results of Cross-Sectional Comparisons of PFAS by Firefighter Status

The results of our cross-sectional PFAS comparisons among firefighters are shown in Tables 3 and 4. Table 3 presents the findings of multiple linear regression analyses comparing incumbent firefighters to recruits. The analyses revealed that incumbent firefighters had significantly lower serum concentrations of n-PFOA and PFNA, at 29% difference and 35% differences, respectively. We did not find any statistically significant differences between incumbent and recruit firefighters in serum PFAS detectability in the models of Sb-PFOA, MeFOSAA, or PFUnDA, as shown in Table 4. Female firefighters exhibited lower concentrations of PFHxS, n-PFOS, Sm-PFOS, n-PFOA, and PFNA compared to their male counterparts (results not shown). Significant positive associations between age and serum concentrations of n-PFOS, Sm-PFOS, and PFUnDA were observed, although these specific findings were not included in Table 3. The sensitivity analyses focusing solely on male firefighters were consistent with the previous results: incumbent status was negatively and significantly associated with n-PFOA and PFNA levels, no statistically significant associations with firefighter status were found for other PFAS compounds, and age was significantly associated with n-PFOS, Sm-PFOS, and PFHxS (the only inferential difference). (Supplementary Tables 2 and 3).

Table 3.

Multiple linear regression analyses of log-transformed serum PFAS concentrations among incumbent (n = 95) and recruit (n = 55) firefighters at study enrollment.

Serum PFAS	Status: Incumbent β^ (95% CI)	Sex: Female β^ (95% CI)
PFHxS	−0.080 (−0.338, 0.178)	−0.890 (−1.155, −0.624) ***
n-PFOS	−0.134 (−0.324, 0.056)	−0.586 (−0.781, −0.390) ***
Sm-PFOS	−0.092 (−0.291, 0.106)	−0.762 (−0.966, −0.557) ***
n-PFOA	−0.339 (−0.530, −0.149) ***	−0.265 (−0.461, −0.069) **
PFNA	−0.433 (−0.669, −0.198) ***	−0.309 (−0.552, −0.066) *
PFDA	−0.143 (−0.311, 0.025)	−0.132 (−0.305, 0.040)

Note: Estimated coefficients (β^) and their corresponding 95% Confidence Intervals (CI) are presented for each PFAS compound. In these models, "recruit firefighters" and "males" are used as reference groups for the "Status" and "Sex" variables, respectively. The β^ values indicate the difference in log-transformed serum PFAS concentrations – expressed in log(ng/mL) – between the specified groups and their corresponding reference categories. A negative β^ value suggests a lower concentration in the group when compared to its reference. Results are denoted as statistically significant at alpha = 0.05 with the following markers: * for p < 0.05, ** for p < 0.01, and *** for p < 0.001. All models have been further adjusted for ethnicity and age.

Table 4.

Odds Ratios from multiple logistic regression analyses for detectable serum PFAS in incumbent (n = 95) and recruit (n = 55) firefighters at study enrollment.

PFAS Detectability	Status: Incumbent [OR (95% CI)]	Sex: Female [OR (95% CI)]
Sb-PFOA	2.513 (0.964, 6.553)	1.833 (0.727, 4.619)
MeFOSAA	0.608 (0.2523, 1.464)	1.015 (0.410, 2.511)
PFUnDA	0.544 (0.217, 1.366)	0.864 (0.329, 2.270)

Note: This table presents Odds Ratios (OR) and their associated 95% Confidence Intervals (CI) for detectable levels of each specified PFAS compound. In these analyses, the reference groups for the "Status" and "Sex" variables are "recruit firefighters" and "males", respectively. The OR values represent the relative likelihood of having detectable serum concentrations of the given PFAS compounds (Sb-PFOA, MeFOSAA, or PFUnDA) at study enrollment. All models have been further adjusted for ethnicity, sex, and age.

Results for Longitudinal Assessment of Changes in PFAS Concentration Over Time

The results for longitudinal assessment of changes in PFAS concentrations over time among male recruit firefighters are presented in Tables 5 and 6. Table 5 shows the results for the linear mixed effects models, indicating that the concentrations of n-PFOS, Sm-PFOS, n-PFOA, PFNA, and PFDA significantly decreased over time. However, there was not a statistically significant relationship between serum concentrations of PFHxS and time. Table 6 presents the results of the association between time since enrollment and serum PFAS detection frequency among male recruit firefighters. The odds of detecting Sb-PFOA were significantly higher at follow-up. In contrast, the odds of detecting MeFOSAA significantly decreased at follow-up. However, there were no significant changes in the frequency of detecting PFUnDA concentrations from enrollment to follow-up.

Table 5.

Associations of time with log-transformed serum PFAS among male recruit firefighters (n = 52).

Serum PFAS	Time (Years) β^ (95% CI)
PFHxS	−0.011 (−0.037, 0.016)
n-PFOS	−0.100 (−0.118, −0.082) ***
Sm-PFOS	−0.094 (−0.112, −0.075) ***
n-PFOA	−0.109 (−0.129, −0.090) ***
PFNA	−0.178 (−0.249, −0.108) ***
PFDA	−0.068 (−0.102, −0.034) ***

Note: Model estimates (β^) and 95% confidence intervals (CI) are reported for time (years) in the six models. These models were exclusively fit to data from male participants due to the limited number of female recruits. Each model includes two measurements per subject and a subject-specific random intercept, as well as fixed effects for years since enrollment, ethnicity, and age. The model estimates represent the change in log-transformed serum PFAS concentrations [log(ng/mL)] over one year. Statistically significant results at alpha = 0.05 are denoted with * for p < 0.05, ** for p < 0.01, and *** for p < 0.001.

Table 6.

Associations of time between study enrollment and follow-up periods with serum PFAS detection frequency among male recruit firefighters (n = 52).

PFAS Detectability	Status: Incumbent [OR (95% CI)]
Sb-PFOA	1.540 (1.100, 2.157) *
MeFOSAA	0.759 (0.578, 0.996) *
PFUnDA	0.934 (0.733, 1.191)

Note: Odds ratios (ORs) and 95% confidence intervals (CIs) are reported for the association between time (years) since enrollment and the frequency of detectable serum Sb-PFOA, MeFOSAA, or PFUnDA concentrations among male recruit firefighters. These models were exclusively fit to data from male participants due to the limited number of female recruits. PFAS concentrations were dichotomized as below the limit of detection. The models were adjusted for ethnicity, age, and time since enrollment. Odds ratios represent the odds of having detectable serum concentrations at follow-up. Statistically significant results at alpha = 0.05 are denoted with * (p < 0.05), ** (p < 0.01), and *** (p < 0.001).

Results for Associations of Fireground Exposure with PFAS at Follow-Up

For associations between fireground exposures and PFAS serum concentrations among male recruit firefighters, we did not find any significant associations between cumulative or most recent fire hours and continuous PFAS serum concentrations, as shown in Table 7. However, we did find a significant association between the detectability of Sb-PFOA at follow-up and days since most recent fire run and most recent structure fire run among male recruits in both the fully-adjusted and minimally-adjusted models (Table 8). In contrast, neither the minimally nor the fully adjusted models showed significant associations between any of the fireground exposure variables and MeFOSAA or PFUnDA.

Table 7.

Estimates and 95% confidence intervals for fireground exposure variables from models of log-transformed serum PFAS concentrations in male recruit firefighters (n = 52) at follow-up.

PFAS	Model Adjustments	Number of Fire Runs	Duration of Fire Runs (Hours)	Days Since Most Recent Fire Run	Number of Structure Fire Runs	Duration of Structure Fire Runs (Hours)	Days Since Most Recent Structure Fire Run
		β^ (95% CI)	β^ (95% CI)	β^ (95% CI)	β^ (95% CI)	β^ (95% CI)	β^ (95% CI)
PFHxS	Minimal	0.0006 (−0.0037, 0.0049)	0.0041 (−0.0032, 0.0114)	0.0005 (−0.0006, 0.0015)	−0.0026 (−0.0128, 0.0075)	0.0046 (−0.0055, 0.0147)	−0.0001 (−0.0009, 0.0007)
	Full	0.0000 (−0.0045, 0.0044)	0.0032 (−0.0047, 0.0111)	0.0004 (−0.0006, 0.0015)	−0.0047 (−0.0153, 0.0059)	0.0031 (−0.0078, 0.0141)	−0.0001 (−0.0009, 0.0007)
n-PFOS	Minimal	0.0000 (−0.0030, 0.0030)	0.0017 (−0.0034, 0.0067)	−0.0004 (−0.0010, 0.0003)	0.0005 (−0.0067, 0.0076)	0.0048 (−0.0021, 0.0117)	−0.0003 (−0.0008, 0.0003)
	Full	0.0000 (−0.0031, 0.0031)	0.0019 (−0.0037, 0.0074)	−0.0003 (−0.0010, 0.0004)	0.0004 (−0.0071, 0.0080)	0.0054 (−0.0021, 0.0128)	−0.0003 (−0.0008, 0.0003)
Sm-PFOS	Minimal	−0.0001 (−0.0033, 0.0030)	0.0010 (−0.0043, 0.0063)	0.0007 (0.0000, 0.0014)	0.0004 (−0.0071, 0.0078)	0.0034 (−0.0039, 0.0106)	0.0004 (−0.0002, 0.0009)
	Full	−0.0001 (−0.0034, 0.0031)	0.0014 (−0.0043, 0.0071)	0.0006 (−0.0001, 0.0014)	0.0008 (−0.0071, 0.0086)	0.0043 (−0.0035, 0.0121)	0.0003 (−0.0002, 0.0009)
n-PFOA	Minimal	0.0013 (−0.0019, 0.0044)	0.0029 (−0.0025, 0.0083)	−0.0003 (−0.0011, 0.0004)	0.0025 (−0.0050, 0.0099)	0.0055 (−0.0018, 0.0128)	−0.0004 (−0.0010, 0.0002)
	Full	0.0014 (−0.0019, 0.0047)	0.0037 (−0.0021, 0.0095)	−0.0004 (−0.0012, 0.0004)	0.0031 (−0.0048, 0.0109)	0.0069 (−0.0009, 0.0147)	−0.0004 (−0.0010, 0.0001)
PFNA	Minimal	0.0036 (−0.0051, 0.0123)	0.0060 (−0.0088, 0.0208)	0.0000 (−0.0020, 0.0020)	0.0020 (−0.0187, 0.0226)	0.0049 (−0.0154, 0.0253)	−0.0009 (−0.0024, 0.0007)
	Full	0.0057 (−0.0030, 0.0143)	0.0123 (−0.0028, 0.0274)	0.0000 (−0.0020, 0.0019)	0.0078 (−0.0131, 0.0287)	0.0134 (−0.0076, 0.0344)	−0.0008 (−0.0024, 0.0007)
PFDA	Minimal	−0.0005 (−0.0055, 0.0045)	0.0042 (−0.0042, 0.0126)	0.0002 (−0.0009, 0.0013)	0.0013 (−0.0102, 0.0128)	0.0096 (−0.0015, 0.0207)	0.0001 (−0.0008, 0.0010)
	Full	−0.0009 (−0.0061, 0.0043)	0.0035 (−0.0057, 0.0128)	0.0002 (−0.0009, 0.0014)	0.0000 (−0.0122, 0.0121)	0.0091 (−0.0030, 0.0213)	0.0001 (−0.0008, 0.0010)

Note: Model estimates (β^) and corresponding 95% confidence intervals (CI) for the association between PFAS exposure and fireground exposure variables. Two models are reported for each unique combination of PFAS and exposure: a minimally-adjusted and a fully-adjusted model. The minimally-adjusted model includes covariates for years since enrollment and log-transformed PFAS concentration [log(ng/mL)] at enrollment, while the fully-adjusted model further adjusts for age at enrollment and ethnicity. These models were exclusively fit to data from male participants due to the limited number of female recruits. The reported coefficients for exposure variables represent the estimated change in the log-transformed PFAS concentration [log(ng/mL)] associated with a one-unit increase in the exposure, while holding the other predictors constant. Statistically significance at alpha = 0.05 was not observed for any of the results.

Table 8.

Odds ratios and 95% confidence intervals from models of serum PFAS detectability and fireground exposures for recruit firefighters (n = 52) at follow-up.

PFAS	Model Adjustments	Number of Fire Runs	Duration of Fire Runs (Hours)	Days Since Most Recent Fire Run	Number of Structure Fire Runs	Duration of Structure Fire Runs (Hours)	Days Since Most Recent Structure Fire Run
		OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Sb-PFOA	Minimal	1.02 (0.97, 1.06)	1.07 (0.99, 1.17)	1.01* (1.00, 1.03)	1.00 (0.91, 1.11)	1.09 (0.99, 1.22)	1.01* (1.00, 1.02)
	Full	1.01 (0.96, 1.06)	1.06 (0.98, 1.16)	1.01* (1.00, 1.03)	0.98 (0.88, 1.09)	1.08 (0.97, 1.21)	1.01* (1.00, 1.02)
MeFOSAA	Minimal	0.99 (0.94, 1.03)	0.99 (0.91, 1.08)	1.00 (0.99, 1.01)	1.06 (0.95, 1.21)	1.06 (0.95, 1.20)	1.00 (0.99, 1.01)
	Full	0.97 (0.92, 1.02)	0.96 (0.87, 1.05)	1.00 (0.99, 1.01)	1.04 (0.92, 1.18)	1.02 (0.90, 1.16)	1.00 (0.99, 1.01)
PFUnDA	Minimal	1.00 (0.95, 1.05)	0.99 (0.91, 1.09)	1.00 (0.99, 1.01)	0.97 (0.86, 1.10)	1.01 (0.89, 1.14)	1.00 (0.99, 1.01)
	Full	1.00 (0.95, 1.06)	1.00 (0.91, 1.10)	1.00 (0.99, 1.01)	0.97 (0.85, 1.10)	1.02 (0.89, 1.17)	1.00 (0.99, 1.01)

Note: The odds ratios (ORs) and 95% confidence intervals (CIs) presented in this table represent reported the association fireground exposure variables and the likelihood of detecting serum Sb-PFOA, MeFOSAA, or PFUnDA concentrations among male recruit firefighters. Due to the limited number of female recruits, these models were exclusively fit to data from male participants. PFAS concentrations were categorized as detectable (at or above the limit of detection) or non-detectable (below the limit of detection). Minimally-adjusted models include the primary predictor variable of time (years) since enrollment and PFAS detectability at enrollment. Fully-adjusted models are further adjusted for age at enrollment and ethnicity. Statistically significant results at alpha = 0.05 are denoted with * (p < 0.05), ** (p < 0.01), and *** (p < 0.001).

Discussion

This study of a single municipal fire department found that legacy serum PFAS concentrations in incumbent firefighters were either similar to or lower than those of recruit firefighters and that serum PFAS concentrations did not increase over time or in association with cumulative fireground exposures. These findings were contrary to our initial expectation that serum PFAS would increase with firefighting activities.

Previous studies have reported elevated concentrations of serum PFAS in firefighters compared to the general population. However, to the best of our knowledge, the present study is the first to make a direct comparison of serum PFAS concentrations between incumbent firefighters and recruits. It is generally expected that recruits, the great majority of whom were without prior firefighting experience, would have similar serum PFAS concentrations as the general population. Surprisingly, we found that serum PFAS concentrations in incumbents were not significantly higher than those of new recruits, which contrasts with previous studies. Specifically, we reported elevated serum PFHxS concentrations (and lower PFNA and PFUnDA) in firefighters from the Tucson and Phoenix areas, who had blood collected in 2009, compared to the results from the National Health and Nutrition Examination Survey (NHANES).⁵⁶ The geometric mean PFHxS concentration for the Arizona firefighters in 2009 was 3.07 ng/mL, whereas in the current study the geometric mean PFHxS concentration at enrollment (2015 - 2016) was 2.62 ng/mL for incumbents and 2.83 ng/mL for recruits, consistent with evidence of declining concentrations of PFHxS over time in the general population.⁵⁷ These concentrations are still higher than the geometric mean serum PFHxS concentrations of 1.22 ng/mL in the 2015-2016 NHANES.⁵⁸

We found lower serum PFAS levels in incumbent women firefighters as compared with male firefighters. These differences echo prior research suggesting that females may metabolize certain PFAS compounds differently than males. For instance, studies have demonstrated that females can excrete PFAS more effectively during menstruation, pregnancy, and breastfeeding.^42,43 However, the limited number of female recruits restricted our ability to draw confident conclusions about sex-specific results.

Drinking water contamination could be a potential source of PFAS exposure in Arizona firefighters. The US EPA’s Third Unregulated Contaminant Monitoring Rule from 2013-2014 showed that six of eight public water services within Pima County, AZ reported concentrations of PFHxS and PFOS above the Minimum Reporting Level. The presence of PFAS in these public water systems could be attributed to point sources of contamination, such as military training areas, AFFF certified airports, and wastewater treatment plants, which can affect the frequency of PFAS detection in local watersheds.⁵⁹ Unfortunately data on firefighters’ residential or fire station water sources were not collected.

Comparing concentrations in firefighters to those of the general population can help determine the extent to which firefighting as an occupation contributes to serum PFAS concentrations. In a recently published study where we compared serum PFAS concentrations between firefighters from four US fire departments to those of NHANES participants, serum PFHxS, Sm-PFOS, n-PFOS, n-PFOA, and PFNA concentrations were significantly elevated in firefighters from at least two fire departments compared to NHANES.⁶⁰ Although the possibility of environmental exposures cannot be discounted as a contributing factor, numerous studies have consistently reported elevated serum PFAS concentrations among firefighters, indicating that occupational exposure is likely a significant contributor. For instance, a study conducted in Southern California with a group of 101 firefighters found that serum concentrations of PFDA were about three times higher, on average, compared to the general US population. In addition, firefighters who did not have their turnout gear professionally decontaminated in the past 12 months had higher serum concentrations of PFOA and PFNA compared to those who did receive decontamination.⁶¹ Graber et al. (2021) assessed years of firefighting and serum PFAS concentrations among volunteer firefighters (n = 135) in the state of New Jersey and found positive associations between years of firefighting and PFDA and perfluorododecanoate (PFDoA). The same study reported that serum concentrations of PFNA, PFDA, and PFDoA were elevated in volunteer firefighters compared to NHANES.⁴⁴ Shaw et al. (2013) found that concentrations of PFOA and PFNA were about twofold higher in San Francisco, California firefighters (n = 12) when compared to the general US population, whereas PFOS and PFHxS were approximately twofold lower; but this study compared data collected from firefighters in 2009 to general US population data from 2003-2004, a potential limitation based on the previously described temporal changes in PFAS concentrations observed in the general US population.^33,57 Within the mid-Ohio Valley, Jin et al. (2011) found that PFHxS was elevated in male firefighters (n = 36) when compared to concentrations in males from other occupational categories (n = 8826); however no differences were found for PFOS concentrations.⁶² Finally, among plasma samples (n = 458) of first responders to the 2001 World Trade Center (WTC) collapse in New York City, concentrations of PFHxS and PFOA were about twofold higher than the general US population. ³² Additionally, serum concentrations of PFHxS and PFOA were higher in WTC first responders who were exposed to more significant amounts of dust (n = 34) and smoke (n = 144). Despite the variations in specific PFAS and exposure scenarios reported across different studies, the consistent finding of elevated PFAS concentrations among firefighters highlights the importance of occupational exposure as a contributing factor.

The longitudinal analyses in the current study show a general decline in PFAS serum concentrations among male recruits from enrollment to follow-up. Specifically, there was a decline in serum concentrations of n-PFOS, Sm-PFOS, n-PFOA, PFNA, and PFDA. Additionally, the odds of having detectable concentrations of MeFOSAA were significantly lower at follow-up. These findings align with recent NHANES data, which suggest that many legacy PFAS concentrations have been declining in the general US population from 2011 to 2018.⁵⁸ However, the lack of a statistically significant decline in PFHxS and PFUnDA among the new recruits is unclear and may include occupational and environmental exposures as well as differences in elimination half-lives. Retired fluorochemical production workers had a longer elimination half-life for PFHxS than those exposed to contaminated drinking water,^63,64 suggesting that occupational exposure may have contributed to serum PFHxS concentrations among male recruits, though the possibility of environmental exposure cannot be ruled out.

Contrasted with the other PFAS chemicals, the odds of having detectable serum concentrations of Sb-PFOA were significantly higher at follow-up among recruit males. This finding is consistent with our previous report on the increased frequency of detection of Sb-PFOA among NHANES males from 2015-16 to 2017-18, which increased from 2 to 11%.⁶⁵ Therefore although we cannot rule out the possibility that firefighting activities may have contributed to this increase, environmental exposure was likely.

A major focus of the current study was the evaluation of the relationship between fireground exposure and serum PFAS concentrations in recruit firefighters. We found no significant relationship between cumulative fireground exposures and serum PFAS in male recruits. However, an unexpected finding was the slight increase in Sb-PFOA detectability associated with days since most recent fire and days since most recent structural fire. This finding was unexpected because exposure to PFAS in combustion products would be expected to decrease with time since most recent fire. Overall, our study did not find evidence to support an association between fireground exposure and increased serum PFAS in recruit firefighters. These findings suggest that exposure to PFAS during firefighting may not be a significant contributor to chronic serum concentrations of legacy PFAS.

While our study did not find a positive association between fireground exposures and serum PFAS concentrations, other studies have emphasized the importance of PFAS exposures for firefighters. For instance, Peaslee et al. (2020) discovered that turnout gear had comparable PFAS concentrations to those found in fire station dust, with the chemical compounds in dust measurements being more similar to those in turnout gear than in AFFF formulations.³⁰ A study by Levasseur et al. (2020) utilizing silicone wristbands worn by firefighters found that firefighters’ PFOS exposure was 2.5 times higher during on-duty responses to fires.⁶⁶ Although these studies did not directly establish a connection between occupational PFAS exposure and firefighters’ serum concentrations, they suggest that occupational exposure may contribute to the overall PFAS body burden among firefighters. Furthermore, it is plausible that the implementation of practices by the Tucson Fire Department to maintain station and gear cleanliness and gear may could have contributed to the observed reduction in serum PFAS concentrations in this population of firefighters. While the Tucson Fire Department does not frequently use AFFF, it is important to acknowledge that the use of AFFF in other fire departments has been associated with higher serum PFAS concentrations, particularly for PFHxS, PFNA, and PFOS,^28,29

Our study had several limitations. We focused on nine commonly detected PFAS, which represents only a subset of the numerous PFAS that firefighters may encounter during their work. While we assessed firefighters’ exposure to PFAS for cumulative firefighting hours and events, other potential sources of exposure were not assessed. For instance, we did not collect information on live-fire training hours, which may expose firefighters to PFAS through contaminated personal protective equipment (PPE) such as turnout gear. Moreover, firefighters could potentially be exposed to PFAS through direct contact with contaminated dust or PPE during other non-emergency activities. Previous studies have reported the presence of PFAS-contaminated dust in fire stations.^31,67 Our study did not collect data on contaminated food or drinking water sources, which could also be a potential source of exposure. We only examined a single fire department, which may not be generalizable to other fire departments. We could not rule out a potential healthy worker effect when comparing incumbents to recruits. Our number of female recruits was limited, preventing an adequate evaluation of longitudinal changes in this group.

This study also has notable strengths. It is the first to compare serum PFAS concentrations in incumbent and recruit firefighters, and to evaluate longitudinal changes in new recruits through their initial firefighting career. We also examined different fireground exposures to determine the relationship between combustion events and serum PFAS concentration at follow-up, providing valuable information on specific firefighting activities. These findings have the potential to inform future research and policy decisions related to the health and safety of firefighters.

Conclusions

Our findings showed that incumbent firefighters had lower serum concentrations of n-PFOA and PFNA compared to recruits. We also observed declining serum PFAS concentrations or detectability over time for most PFAS compounds among recruit firefighters, except for PFHxS, Sb-PFOA, and PFUnDA. Our results suggest that PFAS in combustion products may not be a significant source of exposure for this group of firefighters, as we did not find clear associations between cumulative exposure to fires and PFAS concentrations.

Learning Outcomes:

• Compare serum PFAS concentrations in incumbent and recruit firefighters.

• Evaluate changes over time in recruit firefighter serum PFAS concentrations and determine the association with cumulative measures of fireground exposure.

Data Availability:

De-identified data can be requested by contacting Dr. Burgess and will be reviewed by a representative of the Tucson Fire Department.