Associations Between Seafood Consumption and Serum PFAS Levels Among Asian/Pacific Islanders in the San Francisco Bay Area, California

Abstract

Background

Prior U.S. studies have reported higher per- and polyfluoroalkyl substances (PFASs) levels among Asian/Pacific Islanders (API) than other race/ethnicity groups. High seafood consumption may disproportionately expose API communities to adverse health effects from PFAS-contaminated seafood. We estimated associations between seafood consumption and serum PFAS levels among Chinese and Vietnamese American adults in the San Francisco Bay Area, California.

Methods

From 2016 to 2017, Biomonitoring California collaborated with community partners to recruit 195 participants. Participants completed an exposure questionnaire and provided blood samples for analysis of PFASs and mercury. We pooled associations between seafood consumption and six PFASs using multivariable linear regressions adjusted for demographic covariates and immigration history using a two-stage approach.

Results

PFOS, PFUnDA, PFDA, and PFNA increased 6–10% per five meals of any seafood over the past 30 days. Consuming fish more than three times per week over the past 30 days was associated with higher PFOS (38%) and PFUnDA (42%) compared to consumption below USDA guidelines of at least two meals of fish per week. Caught fish consumed three or more times per week over the past year was associated with 66–125% higher levels of PFOS, PFUnDA, PFDA, and PFNA, compared to consumption less than once per month. Consumption of non-fillet fish parts was associated with 34–124% higher PFOS, PFUnDA, PFDA, PFNA, PFHxS, and PFOA levels compared to no consumption.

Conclusions

These findings suggest a significant route of PFAS exposure for two populations and identify opportunities to reduce exposures through expanding testing and developing culturally appropriate advisories for seafood consumption.

Supplementary Information

The online version contains supplementary material available at10.1007/s12403-025-00743-y.

Introduction

Per- and polyfluoroalkyl substances (PFASs), a class of thousands of environmental chemical pollutants, have been used widely in industrial processes and consumer products for over six decades. The use and production of some PFASs have declined in the United States since the early 2000s due to voluntary manufacturer phaseouts (EPA 2000) and governmental regulations. However, production of these PFASs has continued in other countries around the world and novel PFASs have emerged. Due to their resistance to degradation, these chemicals have contaminated water systems and accumulated widely in humans and wildlife (Li et al. 2018; Kwiatkowski et al. 2020). Of concern, PFASs have been detected in the blood of nearly all individuals tested in many parts of the world (CDC 2019; Zhou et al. 2019; Bjerregaard-Olesen et al. 2017; Calafat et al. 2019). Emerging evidence has linked even low levels of exposure to a variety of negative health effects, such as disrupted thyroid function (Xing et al. 2024), immune suppression (ATSDR 2021; Hollister et al. 2023) and certain cancers (Goodrich et al. 2022; van Gerwen et al. 2023; Ayodele and Obeng-Gyasi 2024). Identifying modifiable predictors of exposure are critical in informing exposure reduction and mitigation strategies, especially for communities with elevated exposures to PFASs.

Several routes of human exposure to PFASs have been identified, including ingestion of contaminated water and food (Hu et al. 2016; Roth et al. 2020), inhalation of contaminated dust, and dermal contact with cleaning and personal care products (Poothong et al. 2019; Ao et al. 2019; Kang et al. 2016). In communities with drinking water contaminated by PFASs, drinking water is likely the dominating route of exposure (Barton et al. 2020; Johanson et al. 2023). However, growing evidence suggests that in the general population, seafood consumption is a major source of PFAS exposure, as PFASs have been demonstrated to bioaccumulate in fish and shellfish (Miranda et al. 2021; Pickard et al. 2022; George et al. 2023). While fish is an important source of nutrients, fish can accumulate mixtures of pollutants, including mercury, dioxins, polychlorinated biphenyls, and other chemicals. Studies across geographical locations have identified widespread PFAS contamination in both marine and freshwater aquatic species (Soerensen et al. 2023; Chen et al. 2023; Barbo et al. 2023). Efforts from some states to account for PFAS contamination in fish consumption advisories, predominantly focusing on legacy PFASs, have estimated up to 50% of perfluorooctane sulfonic acid (PFOS) exposure coming from seafood (WSDH 2022). Furthermore, in 2024, the U.S. Environmental Protection Agency (EPA) updated fish contaminant monitoring recommendations for states, Tribes, and territories to include PFASs, due to widespread contamination and consequent harmful effects on human health.

Existing studies have linked seafood consumption to elevated serum PFAS levels within the general U.S. population and specific communities such as anglers, but few have examined Asian/Pacific Islander (API) communities, in spite of their higher seafood consumption and PFAS levels (Papadopoulou et al. 2019; Christensen et al. 2017; Zhou et al. 2019). In representative data from the National Health and Nutrition Examination Survey (NHANES), significantly higher seafood consumption has been reported among non-Hispanic Asian adults and youth compared with their Hispanic, non-Hispanic black, and non-Hispanic white counterparts (Terry et al. 2018). Additionally, compared with other race/ethnicities, Asians in the U.S. have had generally elevated levels of serum PFASs, including PFOS, perfluorodecanoic acid (PFDA), and perfluorononanoic acid (PFNA) (Sonnenberg et al. 2023), as well as high PFAS burden scores calculated from the sum of 8 PFASs (Liu et al. 2023). Gaps also remain in characterizing exposures associated with various seafood components, such as fish head, skin, or organs, that are consumed by API and other communities (Shaw et al. 2023; Xu and Newman 2015; Sechena et al. 1999). Of concern, PFAS concentrations can be higher in fish heads and organs than muscle tissue (Xie et al. 2024; Stecconi et al. 2024; Soerensen et al. 2023).

Biomonitoring California aimed to address these gaps by collaborating with local community groups to conduct a study of Chinese and Vietnamese adults in the San Francisco Bay Area, California. Our objective was to characterize PFAS exposures in two vulnerable populations and to evaluate associations between serum PFAS levels and seafood consumption reported in questionnaires. We additionally characterized blood mercury, a well-established proxy for seafood consumption, and examined its correlation with serum PFAS levels.

Methods

Study Design and Population

The study population consisted of 200 participants recruited as part of the Asian/Pacific Islander Community Exposures (ACE) Project, a two-phased biomonitoring study among API sub-populations in California. ACE was conducted in collaboration with two community-based nonprofit organizations. APA Family Support Services (APA), a long-term partner of the California Department of Public Health (CDPH), is a non-profit organization based in San Francisco’s Chinatown and has been a trusted source of health education and support in the API community since 1987. The Vietnamese Voluntary Foundation (VIVO), founded in 1979, provides comprehensive acculturation, employment, health, and human support to refugees, immigrants, and low-income families in Santa Clara and San Joaquin counties. For this study, trained APA and VIVO staff worked with Biomonitoring California to create study instruments and carry out community engagement, participant recruitment, and sample collection. Additional support was provided by the Asian and Pacific Islander Family Resources Network (APIFRN), a collaboration of community support organizations.

Recruitment was primarily conducted by APA and VIVO, with additional outreach to other immigrant support groups, dissemination of flyers, and word of mouth between participants. ACE 1 (2016) recruited 100 Chinese American participants in the San Francisco Bay area in collaboration with APA. ACE 2 (2017) recruited 100 Vietnamese Americans primarily in the San Jose area in collaboration with VIVO. To be eligible, study participants were required to be 18 years or older, be of Chinese or Vietnamese descent, have lived in the area for at least 1 year prior to sample collection, and not have traveled outside of the U.S. within the past 30 days. Enrolled participants completed an exposure questionnaire and provided blood samples for biomonitoring. While sample collection and questionnaire administration occurred on the same day for the majority of participants in ACE 2, they occurred on average 11 days apart for 67% of ACE 1 participants, due to phlebotomist scheduling issues and participant availability. Furthermore, four participants from ACE 1 and one participant from ACE 2 were not able to provide blood and were excluded from this study.

All written materials and communications were available in English, Chinese, and Vietnamese, and interpretation services were available. Each participant was provided $100 in cash as a project incentive, as recommended by the community partners.

Biomonitoring California’s standard practice is to make individual results available to all study participants. At the conclusion of the study, participants who requested results (n = 197) were sent paper reports detailing their individual biomarker concentrations. Results return materials also provided comparisons of individual results to overall study results, known health effects of the biomarkers, strategies for reducing exposures, and contact information to speak with study staff.

The study was approved by the California Committee for the Protection of Human Subjects. Written informed consent was obtained from all study participants.

Data Collection

Participants enrolled in the study from June 21, 2016 to November 9, 2016 for ACE 1, and from April 11, 2017 to June 10, 2017 for ACE 2. Trained study staff administered the questionnaire in the language of the participant’s choosing or with a language interpreter present. Participants reported age, race/ethnicity, birth country, years lived in the U.S., primary language, educational attainment, income, diet (including intake of fish, shellfish, and seafood products over the past 30 days and past year), and smoking history. For analysis, we combined primary languages reported by participants into “non-English” (Cantonese, Mandarin, Vietnamese, and other) or “English”.

Participants were asked to characterize their fish and shellfish consumption over two time periods relative to sample collection: within the past year and prior 30 days. This distinction was intended to capture typical fish consumption habits over a longer period of time, as there can be seasonal aspects to fishing. The shorter time period of consumption was included to reduce difficulties in participant recall and as a more relevant timeframe for potential mercury exposures. Consumption characteristics assessed over a longer period of time included: consumption of fish parts (any fish parts, head, skin, eyes, and/or organs; ever/never) (introduced only in ACE 2); fish paste, fish cakes, or balls (yes/no); sauces or flavorings containing shrimp or crab (yes/no) in the past year, and consumption of caught fish in the past year (greater than or equal to three times/week, one to two times/week, one to three times/month, and less than once/month). Over the past 30 days, participants were asked about the number of times fish and shellfish, both bought and caught, were consumed. Bought seafood was defined as eating fish or shellfish from stores, restaurants, or street sellers. Caught fish was defined as fish caught by participants, friends, or family.

We derived additional seafood exposure variables covering the prior 30 days including: total fish meals consumed (bought and caught combined), total seafood meals consumed (fish and shellfish combined), and frequency of fish consumption (bought and caught fish more than three times/week, two to three times/week, and less than two times/week). We were not able to evaluate the frequency of caught shellfish consumption over the prior 30 days because of low reported consumption. Fish consumption frequency categories were based on existing USDA dietary guidelines for fish consumption (two to three servings/week or 8–12 oz/week for pregnant individuals, or at least two servings/week for non-pregnant individuals) that have been adopted by the EPA (USEPA 2024) and Food and Drug Administration (FDA), and to align with similar published studies (Bjorke-Monsen et al. 2020; Zhou et al. 2019).

For PFAS analyses, phlebotomists collected blood in 3.5 mL serum separation tubes with clot activators, avoiding Teflon-coated materials. Samples were inverted several times immediately after collection and then placed upright in a rack to allow blood samples to clot at room temperature between 30 min to 2 h. Clotted blood samples were then centrifuged in the field to separate the serum portion. Processed serum samples were frozen in labeled polypropylene (PP) or polyethylene (PE) collection containers, at −15 °C or below and transferred to Environmental Chemistry Laboratory (ECL) in Berkeley, California where samples were stored at −20 °C until analysis. Samples were stored for 34 weeks on average before analysis. The serum extraction/cleanup and analytical method was slightly modified from our earlier studies (Houtz et al. 2016; Smith et al. 2016). Briefly, 0.1 mL of serum was aliquoted and spiked with internal isotopic-labeled standards. Samples were allowed to equilibrate for 1 to 2 h before undergoing methanol extraction. Clean-up of the extract was performed by adding 15–30 mg ENVI-CARB to the extract, mixing the samples on a rotary table for 5 min, and then centrifuging at 3000 rpm for 5 min. The supernatant was transferred to a glass centrifuge tube and evaporated to near dryness under a gentle nitrogen stream. Methanol was then added to the extract for a total volume of 0.15 mL. The sample extract was then transferred to HPLC vials containing 0.15 mL of Milli-Q water. The final extracts were analyzed using liquid chromatography/tandem mass spectrometry (Shimadzu Nexera X2 UPLC system/Sciex API 5500 QTrap) for 32 PFAS: nine perfluoroalkyl carboxylic acids; four perfluoroalkyl sulfonic acids; four telomer acids; two unsaturated telomer acids; three fluorotelomer sulfonic acids; three polyfluorinated phosphate esters; three n-alkylperfluorooctanesulfonamidoacetic acids; two perfluoroalkylphosphinates; and two perfluoroalkylphosphonic acids. Analytes of interest were separated on a C18 UPLC column (Acquity column, Waters Corporation) before entering the MS/MS system (ABSciex API 5500 QTrap) for multiple-reaction-monitoring (MRM) analysis. The area of the Q1/Q3 ion pairs was used in the analysis. Duplicate samples of NIST SRM 1958 (Organic Contaminants in Fortified Human Serum) were analyzed with each batch of samples to verify method accuracy. Duplicates of three in-house QCs (≈ 0.1 ng/mL, ≈ 1 ng/mL, and ≈ 10 ng/mL) containing all compounds of interest were also analyzed with each batch to monitor method recovery and precision. Duplicate blank samples (bovine calf serum) are additionally processed with each batch of samples to monitor background levels. The lab also participates in external Proficiency Testing (PT), using PT samples from the CDC and the Arctic Monitoring and Assessment Program Ring Test for Persistent Organic Pollutants in Human Serum, at least three times per year to monitor method accuracy.

For heavy metal analyses, California-licensed phlebotomists collected 5 mL of whole blood in 6 mL EDTA blood collection tubes (Royal Blue top). Samples were inverted several times immediately after collection and then stored upright at −20 °C in freezers or Credo Cubes until delivery to CDPH Environmental Health Laboratory (EHL) in Richmond, California. EHL stored specimens at −20 °C for no more than two weeks before analysis. Blood samples were analyzed for mercury following the method detailed elsewhere (Choe and Gajek 2016). Briefly, analyses were conducted using inductively coupled plasma mass spectrometry (ICP-MS). The Agilent 7700 ICP-MS was operated with an integrated sample introduction system (ISIS) for flow injection analysis. Polyatomic interferences were minimized by using a helium-mode collision cell. Three levels of QC materials and two external standard reference materials were prepared and analyzed with each analytical run, and all participant samples were analyzed in duplicate.

Statistical Analyses

We restricted our analysis to the following six PFASs with greater than 65% detection frequencies: PFDA, PFHxS, PFNA, PFOA, PFOS, and PFUnDA. PFAS concentrations below the limit of detection (LOD) were substituted with the LOD/√(2). All analyses utilized log-transformed serum PFAS levels due to a right skew observed in the data. Differences in the distribution of potential covariates by ACE phase were examined with Chi-square statistics for categorical variables and two-sided Student’s t-tests for continuous variables. Correlations between serum PFASs and blood mercury levels were examined by Pearsons’ correlations. Since many of the PFASs were highly correlated, each PFAS was considered separately in analyses. In order to compare ACE exposure levels to national levels, geometric means (GMs) of the six serum PFAS concentrations were compared to NHANES 2015–2016 data using two-sided Student’s t-tests, for NHANES participants greater than 20 years old of all races as well as non-Hispanic Asians.

To assess the association of seafood consumption with serum PFAS levels, we used multivariable linear regression models to calculate percent changes and 95% confidence intervals. A two-stage approach was used to estimate pooled percent changes for each PFAS. When models could be run in both ACE phases, the first stage consisted of an estimated ACE phase-specific log percent change, and the second stage pooled the log estimates from both ACE phases, weighted by the inverse of the variance of each phase using a fixed effect model. Estimates were exponentiated to calculate percent change. For continuous exposure variables, estimates were calculated for the median number of meals, or five meals of additional seafood exposure term. When the exposure variable was categorical, we calculated a pooled percent change for each category level separately. The primary analysis was based on participants from ACE 1 and ACE 2 who provided serum blood samples for PFAS analysis. Heterogeneity among the phase-specific percent changes was assessed using the between-studies variance component Q statistic.

As a secondary analysis, we evaluated models within each ACE phase for which seafood consumption data were sufficiently available (n within category > 5). Questionnaire items on fish part consumption were only available in ACE 2. Associations with frequency of past year caught fish consumption were evaluated among ACE 2 participants only due to low numbers among the highest frequency category for ACE 1 participants.

All models were adjusted for age (years), sex (female vs. male), birth country (U.S vs. non-U.S.), educational attainment (no college vs. some college or more), and portion of life in the U.S., and accounted for household clustering using random intercepts for each household. We ran models separately for each PFAS analyte. Portion of life in the U.S. was included as a proxy for differential exposure from body burdens due to different sources and levels in non-U.S. countries. We calculated portion of life by dividing participant age in years by years lived in the U.S. We constructed a directed acyclic graph (DAG) containing the outcome, exposure pathways, and covariates to guide modeling of variables (Supplemental Figure S1). Covariates were considered for inclusion in the model if they acted as predictors of seafood consumption and serum PFAS levels. Previous studies have shown that dietary habits differ by age, sex, race, education, and nativity, and these factors also have been associated with differences in PFAS levels. Income was not included in the models due to missingness in greater than 20% of responses.

The pooled analysis (“meta” package), comparisons of serum PFAS levels by ACE phase, and generation of tables and figures were conducted using R software 4.0.2 (http://cran-r-project.org). Correlations and stratified regression models were analyzed using SAS software (version 9.4; SAS Institute Inc., Cary, NC, USA).

Results

Across all 195 participants who provided serum blood samples for PFAS analysis (ACE 1: 96; ACE 2: 99), slightly over half were female (53%) and the mean age was around 45 years (Table 1). Overall, participants reported ethnicities as Chinese (ACE 1: 96; ACE 2: 1) and Vietnamese (ACE 2: 98). Participants reported birth countries largely within Asia (China: 25%, Hong Kong: 9.2%, Taiwan: 2.1%, Vietnam: 51%, USA: 11%). ACE 2 participants skewed more towards more recent immigrants (Years lived in the U.S.—ACE 1: 20.3 ± 15.4; ACE 2: 16.4 ± 11.6) and primary home languages that were non-English compared with those in ACE 1. Incomes reported in ACE 2 were also generally lower than those in ACE 1 (< $25,000—ACE 1: 27%; ACE 2: 45%). More ACE 1 participants reported completing some college or more (60%) compared to ACE 2 participants (44%). Participants across both phases had median incomes below the state’s median income as reported by the American Community Survey and less representation of individuals with higher educational attainment than seen in the general population (Bureau 2016). In 2016, 71.4% of Chinese Californians and 57.0% of Vietnamese Californians had completed some college or more.

Table 1.

Participant demographics in the ACE project (n = 195)

Characteristic	Overall, N = 195	ACE 1, N = 96	ACE 2, N = 99	p -value a
		N (%)
Sex				0.77
Female	104 (53%)	50 (52%)	54 (55%)
Male	91 (47%)	46 (48%)	45 (45%)
Age (years)	45.6 ± 18.2	43.7 ± 18.8	47.4 ± 17.6	0.16
Ethnicity				< 0.01
Chinese	97 (50%)	96 (100%)	1 (1.0%)
Vietnamese	98 (50%)	0 (0%)	98 (99%)
Primary home language				< 0.01
Cantonese	72 (37%)	70 (73%)	2 (2.0%)
English	23 (12%)	20 (21%)	3 (3.0%)
Mandarin	2 (1.0%)	2 (2.1%)	0 (0%)
Vietnamese	94 (48%)	0 (0%)	94 (95%)
Other	4 (2.1%)	4 (4.2%)	0 (0%)
Birth country				< 0.01
China	48 (25%)	48 (50%)	0 (0%)
Hong Kong	18 (9.2%)	18 (19%)	0 (0%)
Taiwan	4 (2.1%)	4 (4.2%)	0 (0%)
USA	22 (11%)	18 (19%)	4 (4.0%)
Vietnam	99 (51%)	4 (4.2%)	95 (96%)
Other	4 (2.1%)	4 (4.2%)	0 (0%)
Portion of years lived in U.S. (%)	43 ± 30	50 ± 32	36 ± 25	< 0.01
Years Lived in U.S. (years)	18.3 ± 13.7	20.3 ± 15.4	16.4 ± 11.6	0.05
Education attained				0.03
No college	93 (48%)	38 (40%)	55 (56%)
Some college or more	102 (52%)	58 (60%)	44 (44%)
Income level				0.05
< $25,000	71 (36%)	26 (27%)	45 (45%)
$25,001–75,000	65 (33%)	39 (41%)	26 (26%)
> $75,000	22 (11%)	12 (13%)	10 (10%)
Missing	37 (19%)	19 (20%)	18 (18%)

ACE Asian/Pacific Islander Community Exposures

^ap-values testing for differences between ACE 1 and ACE 2 groups using t-tests for continuous variables and Fisher’s exact test for categorical variables

Comparisons of serum PFAS and mercury levels between the ACE phases, and to NHANES are presented in Table 2. Body burdens [GM ± standard deviation (SD)] were highest for PFOS (ACE 1: 6.51 ± 9.44; ACE 2: 7.47 ± 9.39), followed by PFOA (ACE 1: 1.41 ± 3.56; ACE 2: 1.69 ± 2.05), PFNA (ACE 1: 0.99 ± 1.22; ACE 2: 1.10 ± 0.71), and PFHxS (ACE 1: 0.77 ± 0.74; ACE 2: 1.29 ± 1.26). We observed higher levels of mercury, PFHxS, and PFOA among ACE 2 participants compared with those in ACE 1. Serum levels of several PFASs were higher among ACE participants compared with national estimates: PFNA and PFDA were higher among both ACE 1 and ACE 2 participants compared with all adults and Asian adults from the 2015–2016 NHANES, while PFOS was higher among ACE 2 participants only compared with both national estimates. Blood mercury levels were higher among both ACE 1 and ACE 2 participants compared to both national estimates. Among the other 26 PFASs measured, 11 PFASs, including PFAS precursors, short chain PFASs, and other long chain PFCAs, were detected in at least one of the ACE phases, with detection frequencies ranging from 2 to 98% (Supplemental Table 1).

Table 2.

Serum PFAS and blood mercury levels (µg/L) among participants in the ACE project (n = 195)

Analyte	ACE Phase	n	DF (%)	MDL	GM	SD	25th	50th	75th	95th	p-valuea	NHANES All Adults GM b	NHANES All adults p-valuec	NHANES Asians GM d	NHANES Asians p-valuee
PFHxS	ACE 1	96	100	0.009	0.77	0.74	0.15	0.79	2.19	4.33	< 0.01	1.22	< 0.01	0.95	< 0.01
	ACE 2	99	100	0.010	1.29	1.26	0.34	1.21	3.88	9.47			0.38		< 0.01
PFOA	ACE 1	96	100	0.050	1.41	3.56	0.25	1.35	3.18	35.50	0.04	1.60	0.06	1.66	0.02
	ACE 2	99	100	0.060	1.69	2.05	0.25	1.61	3.82	19.90			0.34		0.78
PFOS	ACE 1	96	100	0.057	6.51	9.44	0.84	6.04	26.90	56.30	0.24	5.02	< 0.01	5.75	0.15
	ACE 2	99	100	0.191	7.47	9.39	1.53	7.00	34.22	43.20			< 0.01		< 0.01
PFNA	ACE 1	96	99.0	0.250	0.99	1.22	0.18	0.95	3.12	10.30	0.20	0.59	< 0.01	0.77	< 0.01
	ACE 2	99	99.0	0.250	1.10	0.71	0.18	1.08	2.64	4.33			< 0.01		< 0.01
PFDA	ACE 1	96	80.2	0.250	0.48	0.95	0.18	0.45	2.19	7.80	0.13	0.16	< 0.01	0.31	< 0.01
	ACE 2	99	87.9	0.250	0.56	0.50	0.18	0.54	1.85	2.74			< 0.01		< 0.01
PFUnDA	ACE 1	96	100	0.050	0.40	0.30	0.06	0.43	1.06	1.70	0.20	–	–	0.23	< 0.01
	ACE 2	99	98.0	0.050	0.45	0.45	0.04	0.44	1.18	2.91			–		< 0.01
Hg	ACE 1	96	99.0	0.150	3.58	3.33	0.23	4.12	9.95	18.00	0.01	0.81	< 0.01	1.73	< 0.01
	ACE 2	99	100	0.075	4.77	3.57	0.21	5.18	12.68	17.80			< 0.01		< 0.01

ACE Asian/Pacific Islander Community Exposures, DF detection frequency, MDL method detection limit, Min minimum, GM geometric mean, SD standard deviation, PFDA perfluorodecanoic acid, PFHxS perfluorohexane sulfonate, PFNA perfluorononanoic acid, PFOA perfluorooctanoic Acid, PFOS perfluorooctane sulfonic acid, PFUnDA perfluoroundecanoic acid, Hg mercury

^a p-value from the two-sided Student’s t-test for differences in geometric means by ACE 1 and ACE 2 groups

^b NHANES All Adults 2015–2016 geometric mean data. PFUnDA data not reported due to low detection frequency

^c p-value from the two-sided Student’s t-test for differences in geometric means between ACE group and NHANES data

^d NHANES Non-Hispanic Asian Adults 2015–2016 geometric mean data

A greater proportion of ACE participants (59%) reported consuming any seafood at least twice a week (including both bought and caught fish and shellfish meals) compared with national estimates of all adults (19.2%) in the 2015–2016 NHANES (Terry et al. 2018). Seafood consumption among non-Hispanic Asian adults from the 2013–2016 NHANES was also lower (41%) compared with that of ACE participants. Among ACE participants, more participants reported consuming bought fish compared with fish caught by participants, friends, or family (Table 3). Across all participants, the mean frequency of fish consumption (mean ± SD) in the prior 30 days was 7.1 ± 6.3 times for bought fish compared with 0.4 ± 2.0 times for caught fish. Among those who reported consuming caught fish (32% of all participants), a small portion reported frequent consumption within the past year: 11% consuming once or twice per week and 4% consuming greater than or equal to three times per week. Consumption of caught fish in the past year was slightly higher among ACE 2 participants (37%) compared with those in ACE 1 (27%). An average of 3.6 ± 3.7 meals of bought shellfish was reported in the prior 30 days. Lastly, non-fillet fish part consumption (fish head, skin, organs, and eyes) is reported only for ACE 2 participants, since these questionnaire items were only introduced in the second phase of the ACE Project. The majority of ACE 2 participants (84%) reported ever consuming non-fillet fish parts. When examined by specific fish part, fish head (64%) and skin (76%) were consumed by more participants than fish organs (13%) and fish eyes (34%).

Table 3.

Distributions of seafood consumption in the ACE project (n = 195)

Characteristic	Overall N = 195	ACE 1 N = 96	ACE 2 N = 99	p -value a
	Mean ± SD (number of meals)
Past 30 day total seafood	11.2 ± 8.8	12.7 ± 9.5	9.7 ± 7.9	0.02
Missing, n (%)	6 (3.1%)	4 (4.2%)	2 (2.0%)
Past 30 day total fish	7.5 ± 7.050	8.6 ± 7.6	6.4 ± 6.2	0.04
Past 30 day bought fish	7.1 ± 6.3	8.1 ± 6.7	6.1 ± 5.7	0.02
Past 30 day caught fish	0.4 ± 2.0	0.5 ± 2.2	0.4 ± 1.7	0.76
Past 30 day bought shellfish	3.6 ± 3.7	4.0 ± 3.7	3.1 ± 3.5	0.08
Missing, n (%)	3 (1.5%)	1 (1.0%)	2 (2.0%)
	N (%)
Past Year Caught Fish
No	132 (68%)	70 (73%)	62 (63%)	0.13
Yes	63 (32%)	26 (27%)	37 (37%)
Past Year Frequency of Caught Fish
≥3/week	8 (4.1%)	1 (1.0%)	7 (7.1%)	0.04
1–2/week	22 (11%)	7 (7.3%)	15 (15%)
1–3/month	33 (17%)	18 (19%)	15 (15%)
Less than 1/month	132 (68%)	70 (73%)	62 (63%)
Fish paste, fish cake, or fish balls
No	29 (15%)	10 (10%)	19 (19%)	0.11
Yes	165 (85%)	85 (89%)	80 (81%)
Missing	1 (0.5%)	1 (1.0%)	0 (0%)
Shrimp or crab sauce
No	41 (21%)	26 (27%)	15 (15%)	0.03
Yes	149 (76%)	66 (69%)	83 (84%)
Missing	5 (2.6%)	4 (4.2%)	1 (1.0%)
Any Fish Part b	–	–		–
No			16 (16%)
Yes			83 (84%)
Fish organs b	–	–		–
No			86 (87%)
Yes			13 (13%)
Fish head b	–	–		–
No			36 (36%)
Yes			63 (64%)
Fish skin b	–	–		–
No			24 (24%)
Yes			75 (76%)
Fish eyes b	–	–		–
No			64 (65%)
Yes			34 (34%)
Missing			1 (1.0%)

ACE Asian/Pacific Islander Community Exposures

^a p-values testing for differences between ACE 1 and ACE 2 groups using t-tests for continuous variables and Fisher’s exact test for categorical variables

^b Information collected in ACE 2 only

Statistically significant positive correlations were observed between all the serum PFAS levels, with Rho values ranging from 0.27 (between PFHxS and PFUnDA) to 0.82 (between PFNA and PFDA) (p < 0.01) (Table 4). Correlations with PFOS ranged from 0.37 (for PFOA, which has not been reported to accumulate significantly in fish) up to 0.76 (for PFNA) and 0.79 (with PFDA), two longer chain carboxylic acids that have been commonly detected in fish samples. Blood mercury was significantly and positively correlated with all 6 PFASs: most strongly with serum PFUnDA (Rho = 0.70, p < 0.01) and least strongly with PFOA and PFHxS (0.22 and 0.24, respectively). More moderate associations were detected with PFDA, PFNA, and PFOS, ranging from 0.40 to 0.46 (p < 0.01).

Table 4.

Pearson’s correlation^a between serum PFAS levels and blood mercury among 195 ACE participants

Analyte	PFHxS	PFOA	PFOS	PFNA	PFDA	PFUnDA	Hg
PFHxS	–	0.66	0.52	0.56	0.36	0.27	0.24
PFOA	–	–	0.37	0.67	0.44	0.36	0.22
PFOS	–	–	–	0.76	0.79	0.59	0.40
PFNA	–	–	–	–	0.82	0.69	0.46
PFDA	–	–	–	–	–	0.80	0.46
PFUnDA	–	–	–	–	–	–	0.70
Hg	–	–	–	–	–	–	–

ACE Asian/Pacific Islander Community Exposures, PFDA perfluorodecanoic acid, PFHxS perfluorohexane sulfonate, PFNA perfluorononanoic acid, PFOA perfluorooctanoic Acid, PFOS perfluorooctane sulfonic acid, PFUnDA perfluoroundecanoic acid

^a Pearson’s rho calculated using log-transformed PFAS and mercury values. All correlations were significant at p < 0.01

Pooled adjusted percent change in serum PFAS levels associated with increased seafood consumption over the past 30 days are presented in Table 5. Consumption of seafood five additional times in the past 30 days was associated with 10% higher PFOS (95% CI 4%, 15%), 10% higher PFUnDA (95% CI 4%, 16%), 7% higher PFDA (95% CI 2%, 13%), and 6% higher PFNA (95% CI 1%, 10%). Consumption of five additional meals of bought fish in the past 30 days was associated with higher levels of four PFASs, ranging from 7 to 14% higher levels of PFOS, PFNA, PFDA, and PFunDA, while caught fish consumption was associated with higher levels of all six PFASs. Five additional meals of caught fish in the past 30 days was associated with 39% higher PFOS (95% CI 13%, 63%), 29% higher PFUnDA (95% CI 10%, 51%), 37% higher PFDA (95% CI 10%, 51%), 32% higher PFNA (95% CI 17%, 49%), 16% higher PFHxS (95% CI 5%, 28%), and 15% higher PFOA (95% CI 4%, 27%). When combining both fish parameters into a single variable measuring the total times fish was consumed in the past 30 days (including bought and caught fish), associations were attenuated for PFDA and PFNA and no longer observed for PFHxS and PFOA. No associations were observed for bought shellfish among the pooled estimates. There was evidence of between-study heterogeneity, largely for parameters associated with PFDA and PFHxS.

Table 5.

Adjusted^a pooled percent change (95% confidence interval) in serum PFAS levels per five increment increase or seafood consumption parameter among ACE phases, California, 2016–2017 (n = 195)

Parameter	N	PFHxS		PFOA		PFOS		PFNA		PFDA		PFUnDA
		Percent change (95% CI)	p-valueb	Percent change (95% CI)	p-valueb	Percent change (95% CI)	p-valueb	Percent change (95% CI)	p-valueb	Percent change (95% CI)	p-valueb	Percent change (95% CI)	p-valueb
Continuous
Past 30 Day Total Seafood c	189	2 (−2, 6)	0.03	2 (−2, 7)	0.42	10 (4, 15)	0.45	6 (1, 10)	0.62	7 (2, 13)	0.01	10 (4, 16)	0.25
Past 30 Day Total Fish c	195	1 (−4, 7)	0.05	3 (−3, 9)	0.93	13 (6, 20)	0.45	8 (3, 14)	0.76	9 (−1, 19)	0.82	14 (7, 21)	0.49
Past 30 Day Bought Fish c	195	0 (−6, 6)	0.06	2 (−4, 9)	0.99	12 (4, 20)	0.83	7 (1, 14)	0.64	8 (1, 17)	0.67	14 (6, 23)	0.47
Past 30 Day Caught Fish c	195	16 (5, 28)	0.31	15 (4, 27)	0.75	39 (19, 63)	0.53	32 (17, 49)	0.73	37 (14, 63)	0.29	29 (10, 51)	0.91
Past 30 day Bought Shellfish c	192	5 (−5, 14)	0.39	0 (−9, 9)	0.33	8 (−3, 19)	0.19	2 (−6, 11)	0.31	7 (−2, 16)	0.01	9 (−4, 23)	0.10
Categorical
Past 30 Day Frequency of Total Fish Consumption	195
< 2/week	138	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–
2–3/week	26	−11 (−30, 15)	0.53	−7 (−28, 20)	0.24	6 (−21, 43)	0.29	11 (−11, 39)	0.52	3 (−20, 34)	0.10	17 (−10, 52)	0.06
> 3/week	31	11 (−10, 37)	0.07	8 (−13, 33)	0.81	38 (8, 77)	0.34	18 (−2, 43)	0.32	23 (−4, 57)	0.31	42 (11, 82)	0.35
Past Year Caught Fish	195
No	132	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–
Yes	63	11 (−4, 30)	0.04	15 (−2, 36)	0.17	15 (−8, 43)	0.23	17 (−2, 39)	0.13	27 (2, 58)	0.55	26 (2, 55)	0.43
Fish Paste, Balls, and Cakes	195
No	29	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–
Yes	165	−1 (−23, 28)	0.58	18 (−10, 55)	0.79	21 (−11, 65)	0.29	12 (−15, 46)	0.31	20 (−9, 59)	0.26	40 (−2, 100)	0.52
Shrimp or Crab Sauce	190
No	41	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–	Reference	–
Yes	149	18 (−5, 46)	0.38	12 (−12, 42)	0.25	39 (8, 79)	0.35	25 (0, 56)	0.13	26 (−6, 67)	0.38	29 (−2, 71)	0.36

ACE Asian/Pacific Islander Community Exposures, PFDA perfluorodecanoic acid, PFHxS perfluorohexane sulfonate, PFNA perfluorononanoic acid, PFOA perfluorooctanoic Acid, PFOS perfluorooctane sulfonic acid, PFUnDA perfluoroundecanoic acid

^a Adjusted for age (years), sex, birth country (U.S vs. non-U.S.), educational attainment (no college vs. some college or more), and portion of life in the U.S. (years), and random intercepts for each household

^b p-value for between-study heterogeneity (Q statistic)

^c Estimates represent proportional change in geometric mean of serum PFAS for each five meal increase in fish/shellfish consumed, adjusted for other variables in the model

We examined fish consumption frequency over both the past 30 days and year and compared PFAS levels by category. Participants who had consumed fish, either bought or caught, more than three times per week in the past 30 days had 38% higher PFOS (95% CI 8%, 77%) and 42% higher PFUnDA (95% CI 11%, 82%), compared to those who consumed fish less than twice per week, or below the USDA dietary recommendations (Table 5). We observed limited evidence of between-study heterogeneity, largely for PFDA and PFHxS. When examining caught fish only, participants who reported consuming any caught fish in the past year had 26% higher levels of serum PFUnDA (95% CI 2%, 55%) and 27% higher PFDA (95% CI 2%, 58%). Participants were asked separately about frequency of past year caught fish consumption. Due to low numbers among ACE 1 participants, associations were evaluated among ACE 2 participants only (Fig. 1). Those who consumed caught fish most frequently in the past year (three meals or greater per week) had higher serum levels compared with those who consumed less than one caught fish meal per month: 92% higher PFOS (95% CI 29%, 186%) and for perfluorocarboxylic acids (PFCAs) ranging from C9 to C11, from 66% higher PFNA (95% CI 32%, 108%) to 125% higher PFDA (95% CI 73%, 194%).

Fig. 1.

Percent adjusted change (95% confidence intervals) in each serum PFAS among participants who reported past year caught fish consumption in varying frequency categories compared with those who consumed caught fish less than once per month in the past year (n = 99). Adjusted for age (years), sex, birth country (U.S vs. non-U.S.), educational attainment (no college vs. some college or more), and portion of life in the U.S. (years), and accounted for clustering with random intercepts for each household. Note: PFASs, per- and polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonate; PFOA, perfluorooctanoic Acid; PFOS, perfluorooctane sulfonic acid; PFNA, perfluorononanoic acid; PFDA, perfluorodecanoic acid; PFUnDA, perfluoroundecanoic acid

Consumption of any non-fillet fish part was a significant predictor in all adjusted models for all serum PFAS levels except for PFHxS (Fig. 2). Adjusted percent changes ranged from 34% higher serum PFOA levels to 124% higher PFOS levels. When examining fish parts separately, consumption of fish heads was found to be significant in all adjusted models for PFOS, PFDA, PFNA, PFUnDA, and PFHxS, while consumption of fish eyes was a significant predictor in all adjusted models for PFOS, PFDA, PFNA, and PFUnDA. Consumption of both fish skin and fish organs was associated with higher serum PFOS, PFUnDA, and PFDA levels. We observed moderate associations for some seafood products: shrimp or crab sauce consumption was associated with higher serum levels of PFOS and PFUnDA, while consumption of fish paste, balls, or cakes was not a significant predictor (Table 5).

Fig. 2.

Percent adjusted change (95% confidence intervals) in each serum PFAS among participants who ever consumed non-fillet fish parts compared with those who did not ever consume non-fillet fish parts (any fish part, fish eyes, head, organs, or skin) (n = 99). Adjusted for age (years), sex, birth country (U.S vs. non-U.S.), educational attainment (no college vs. some college or more), and portion of life in the U.S. (years), and accounted for clustering with random intercepts for each household. Note: PFASs, per- and polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonate; PFOA, perfluorooctanoic Acid; PFOS, perfluorooctane sulfonic acid; PFNA, perfluorononanoic acid; PFDA, perfluorodecanoic acid; PFUnDA, perfluoroundecanoic acid

Secondary analyses examining these parameters separately within each ACE phase are presented in Supplemental Table 2. Overall, the pattern observed from the pooled analysis remained, though some results were only significant among ACE 2 participants. These parameters included past 30 day consumption of shellfish, the highest frequency of fish consumption (3 times per week), as well as past year caught fish consumption.

Discussion

In this study, higher fish and shellfish consumption was consistently associated with higher serum levels of several PFASs (PFOS, PFDA, PFNA, and PFUnDA). These associations were strongest for consumption of caught fish and non-fillet fish parts. Participants who consumed fish, including bought and caught fish, more than 3 times per week in the prior 30 days had higher levels of PFOS and PFUnDA compared to those whose consumption fell below the USDA fish dietary recommendations. All six PFASs measured were higher among those who consumed more than 3 servings of caught fish per week in the past year compared with those who did not consume caught fish. Associations between consumption of any non-fillet fish part and serum PFAS levels were observed for five PFASs (PFHxS, PFOS, PFNA, PFDA, PFUnDA) at even larger magnitudes than observed with fillet consumption. Moderate to strong correlations between blood mercury and serum PFASs were observed for PFNA, PFDA, and PFUnDA.

Our findings were consistent with a growing body of evidence that has observed associations with both fish and shellfish consumption and elevated PFAS levels. Studies within the general population across the U.S. (Jain 2014) and California with generally moderate fish consumption (Pennoyer et al. 2024) reported associations with similar PFASs. Jain (2014) examined NHANES participants aged 12 and older from 2003 to 2008 and found positive associations between fish consumption and serum PFAS levels (including PFOS, PFNA, PFUnDA). Christensen et al. (2017) examined NHANES participants in the same age range in 2007–2013 and found elevated PFNA and PFUnDA associated with increased total fish consumption in the past 30 days. They additionally observed associations between shellfish consumption and elevated PFOS, PFNA, PFOA, and PFUnDA. Within California, Pennoyer et al. (Pennoyer et al. 2025) assessed the contributions of diet and drinking water in the California Regional Exposure Study (2018–2020), among a general adult population with generally lower seafood consumption and lower PFAS serum levels compared to ACE participants. This study found weekly seafood intake to be associated with increased PFNA, PFDA, and PFUnDA levels, and drinking water to be associated with PFOA and PFHxS levels. Other studies from outside the U.S. have also found consistent associations with these PFASs and both fish and shellfish consumption (Bjorke-Monsen et al. 2020; Papadopoulou et al. 2019; Zhou et al. 2019; Gyllenhammar et al. 2025).

The observed associations between fish and shellfish consumption and serum PFAS levels are supported by data from sampling efforts demonstrating variable PFAS contamination in seafood samples. Testing from U.S. supermarkets has been limited, based largely on U.S. FDA studies of items frequently consumed by the general U.S. population. While most food items had non-detectable levels of PFAS, evidence of detectable levels of PFASs were found in a subset of seafood samples. These included PFOS in clams from 2010 to 2012 (Young et al. 2013); PFOS, PFNA, and PFDA in frozen fish sticks and canned tuna from the 2018–2019 Total Diet Study analyses (Genualdi et al. 2021); and most recently, PFOA in canned clam samples produced in China from a 2021 targeted seafood study (Young et al. 2022). Notably, higher PFAS concentrations have been measured in non-commercial samples, particularly freshwater samples, a trend that highlights the vulnerability of those who rely on local water bodies for sustenance (Pulster et al. 2022). In similar years as the ACE Project, freshwater fish samples collected in 2013–2015 by the U.S. EPA fish monitoring program had elevated levels of PFOS and longer-chain PFCAs, including PFUnDA, PFDA, and perfluorododecanoic acid (PFDoA) (Barbo et al. 2023). In contrast, shellfish samples have exhibited differences in PFAS accumulation, with shorter chain PFASs, such as PFOA, and perfluorosulfonic acids, such as PFHxS, generally dominating (Koban et al. 2024; Stecconi et al. 2024). In addition, significant variations in tissue PFAS burdens may be present across regions, with higher concentrations near contaminated sites, and differences between and within species in the same body of water (Blazer et al. 2024; Ankley et al. 2021).

Our categorical analyses based on U.S. federal dietary recommendations suggest that fish consumption below the recommended two or greater servings per week was largely associated with lower PFAS levels compared with higher fish consumption (greater than three servings per week). While knowledge gaps remain in understanding seafood consumption contributions to PFAS body burdens, fish has overwhelmingly been recognized as a healthy food, though some factors may threaten this important source of nutrition. This was acknowledged in a 2024 report jointly released by the U.S. FDA and National Academies of Science, Medicine and Science on the role of seafood in child growth and development. The report upheld existing seafood recommendations despite consideration of evidence around PFAS contamination, but current calculations may not capture variation in local fish contamination and higher consumption among some groups, such as Asian, Tribal, and subsistence communities that may be more impacted by PFAS contamination (NASEM 2024).

Few studies have delineated seafood consumption by bought compared to caught sources, the latter of which can be more impacted by local water body contamination. Our study findings indicate the importance of investigating body burdens associated with locally caught fish, as associations with caught fish were stronger compared to bought fish. Other studies focusing on high fishing populations, particularly anglers, in impacted waters have also consistently reported higher PFAS body burdens associated with local seafood consumption (Christensen et al. 2016; Liu et al. 2022; Hölzer et al. 2011). In a sample of Wisconsin male anglers near the Great Lakes, consumption of commercially purchased fish was associated with higher PFHxS levels, while consumption of locally caught fish was associated with higher levels of PFOS, PFOA, PFDA, PFNA, PFUnDA, and perfluoroheptane sulfonic acid (PFHpS) (Christensen et al. 2016). An additional analysis of a statewide representative cohort in Wisconsin from 2014 to 2016 similarly observed an association with several PFASs and past year consumption of caught fish (Pomazal et al. 2024). In France, freshwater fish was similarly linked to increased PFOS, PFNA, PFHxS, PFHpS, and PFDA among an adult angler population (Denys et al. 2014). These studies underscore the need to protect populations with higher consumption of local seafood, including anglers and those practicing subsistence fishing, who may be at higher risk of PFAS exposures and associated health effects.

We found evidence of markedly significant increases in nearly all serum PFASs associated with consumption of fish head, eyes, organs, and skin consumption. These findings indicate that both the quantity of seafood and specific parts consumed contribute to PFAS exposures, yet few biomonitoring studies have examined whole fish or non-fillet fish consumption. One study among Norwegian adults did report larger associations with fish liver consumption than lean fish fillet consumption, for several PFASs, including PFOS, PFUnDA, PFNA, and PFDA (Haug et al. 2010). Studies that have tested fish from across the U.S., including Michigan and Oregon, for PFASs in both fillet and non-fillet parts have frequently detected higher levels of PFASs in non-fillet samples than those in fillet samples, largely based on the liver (Nilsen et al. 2024; Blazer et al. 2024; Capozzi et al. 2023; Figueroa-Munoz et al. 2025). Internationally, similar patterns have been reported in analyses from China (ΣPFAS in viscera 2.18–17.67 compared to 0.45–2.92 ng/g ww in muscle) (Xie et al. 2024) and Italy (ΣPFAS in liver 3.1–10 compared to 0.032–1.7 ug kg^− 1 in muscle) (Stecconi et al. 2024). A temporal analysis on cod (Gadus morhua) livers from 1981 to 2013 demonstrated significant increases in PFASs including PFOS, PFHxS, PFDA, PFUnDA, and PFNA over time (Schultes et al. 2020). While consumption advisories often recommend consuming only the fillet, as shown in this study, existing consumption habits may expose consumers to higher levels of PFASs and other contaminants. Sampling of other non-fillet fish parts, such as head and eyes, may be useful in constructing more inclusive consumption advisories.

Emerging research identifying co-contamination of some fish with mercury and PFASs further supports our identification of fish consumption as a source of exposure to PFASs. Fish and seafood are established dietary sources of mercury and elevated blood mercury levels (Mahaffey et al. 2009; Nielsen et al. 2014), due to accumulation of methylmercury in fish and shellfish. Fish samples have also been seen to have co-contamination of methylmercury and PFASs, particularly those with carbon chains eight or longer (Ahrens et al. 2011; Arinaitwe et al. 2020; Lee et al. 2024). In other biomonitoring studies measuring serum and hair samples, significant correlations between blood mercury and levels of long chain PFASs have been reported, most strongly for PFUnDA and more moderately for PFOS, PFDA, and PFNA (Bjorke-Monsen et al. 2020). Future studies evaluating contaminant exposures associated with seafood consumption may consider impacts of pollutant mixture exposures on health and pollutant elimination.

A key element of our study design was biomonitoring API adults, with the aim to provide insight into two potentially highly exposed populations. All participants in ACE 1 identified as Chinese and nearly all in ACE 2 identified as Vietnamese. The differences seen in ACE, along with data from the Measuring Analytes in Maternal Archived Samples (MAMAS) Study (Biomonitoring California 2024), suggests that aggregating Asian subpopulations into one category may be masking differences in exposure to PFASs. Few studies in the U.S. have disaggregated Asian sub-populations, aside from studies of Chinese, Hmong and Karen women refugees in the Great Lakes area (Shaw et al. 2023). While the small sample size of this study limited our investigation of all seafood consumption parameters for each ACE phase separately, our results have indicated some general differences in both consumption patterns and serum PFAS levels between these two groups. Several factors may explain these PFAS-related differences. The variation in participants’ geographic locations may have contributed to differences in PFAS exposures from drinking water sources, as data from 2013 to 2025 indicate no detectable PFASs in San Francisco water in contrast to detectable levels of PFHxS and PFOS in San Jose (where ACE 2 participants primarily resided), though not for PFNA, PFDA, and PFUnDA (EPA 2025). Body burdens retained from time lived in other countries, and potential differences in PFASs used in each country, could additionally contribute to varying PFAS profiles and body burdens. A larger proportion of ACE 2 participants, who generally had higher levels of PFASs, were born outside of the U.S., and ACE 2 participants had also spent a larger portion of their lives outside of the U.S. compared with ACE 1. However, more studies are needed to further understand heterogeneities within Asian subpopulations, as existing biomonitoring data on Chinese and Vietnamese adult PFAS levels have utilized different PFAS analytical methods and study years limiting cross comparison. Elevated PFAS levels among API adults in the U.S. may be particularly an issue of concern in California due to its large API population.

Our findings further underscore the importance of monitoring PFAS concentrations in seafood throughout the San Francisco Bay Area and communicating relevant consumption advice. The 2024 updates to U.S. EPA recommendations for fish monitoring have recently been expanded to include five PFASs, including PFOS, PFNA, PFDA, PFOA, and PFHxS, due to their known health effects. An additional list of contaminants to watch for, distinguished as those without established federal oral toxicity or reference doses, included seven additional PFASs, including PFUnDA, which was consistently associated with several measures of seafood consumption in this analysis. In California, existing fish consumption advisories issued by the Office of Environmental Health Hazard Assessment are based on mercury, PCBs, and PBDEs with plans to address PFASs in the future (OEHHA). The OEHHA San Francisco Bay fish consumption advisory, based on mercury and PCBs, restricts or recommends against consumption of a number of species. Growing evidence linking fish consumption with higher PFAS levels, including the results of the present analysis, suggest that interventions are needed to address detrimental health impacts associated with PFAS exposures and disparities in these exposures. Such evidence includes local sampling efforts by the San Francisco Regional Monitoring Program, conducted between 2009 and 2019, that demonstrate PFAS contamination in SF Bay Area sportfish samples. PFAS detections largely mirrored ACE findings, with PFOS dominating followed by long chain PFCAs, though the near universal detection of precursors such as 7:3 FTCA deviated from ACE participant PFAS levels (Méndez 2025). Méndez noted levels of PFOS detected in most samples would trigger recommendations for limited fish consumption by Massachusetts, a state with the strictest PFAS-based fish consumption advisories in the U.S. However, sampling in Californian fish samples has been limited primarily to muscle tissue. More data is needed on non-fillet fish parts and across more seafood species, particularly shellfish, to develop appropriate and targeted interventions that may be more health protective than existing advisories.

Several limitations are worth noting in the present analysis. Use of self-reported dietary recall questionnaires may have introduced exposure misclassification. In addition, while the questionnaire did enable us to examine several measures of seafood consumption, portion sizes were not ascertained due to length considerations of the exposure questionnaire. A prior study has found serving sizes among Californian consumers of caught fish to be around 8 ounces, which suggests that we may have underestimated associations with fish consumption in relation to USDA dietary guidelines (Silver et al. 2007). The questionnaires did not collect information on potentially confounding factors that have been found to be associated with PFAS concentrations: drinking water sources, dust, dermal contact sources, parity, geographic location of consumed fish, and seasonal timing of fish consumption. However, use of the stratified secondary analysis helps address differences in drinking water PFAS levels between San Francisco, where no detections were observed, and San Jose residents, where some low level PFHxS and PFOS detections were observed (EPA 2025). Further, the collection of certain information, such as consumption of fish parts, was introduced in the second phase of the study and reduced the sample size for the related analyses. Additional information on how frequently fish parts were consumed over particular time period would be valuable for future studies. While we may expect exposures over the past 30 days and past year to be reasonably similar to longer periods of consumption history, gaps remain in characterizing associations with longer term seafood consumption, as PFASs may have years-long half-lives in humans. Future studies are needed to better understand potential differences in exposure from caught shellfish and differences within species, trophic levels, and habitat, which could not be assessed in this study due to the small sample size. This study, based on a convenience sample of API adults in two regional locations, may not be generalizable to broader populations. Both the Chinese and Vietnamese American adults had PFAS concentrations and fish consumption frequencies generally higher than the general U.S. population. However, a recent report from the EPA suggests that many other ethnic/racial groups in the U.S. consume non-fillet fish parts. Thus, frequent seafood and non-fillet fish part consumers, regardless of racial/ethnic background, may be highly exposed to PFASs (U.S. Environmental Protection Agency 2024).

This study had several strengths. First, the present analysis focused on an understudied population of API individuals living in the U.S. Additionally, the dietary questionnaires provided specific insight on fish and seafood consumption, including detailed consumption habits. The information collected allowed for estimation of associations between PFASs and consumption of specific fish parts such as fish head, eyes, skin, and organs, which has not been evaluated in the majority of existing studies. In contrast to most previous analyses that analyzed only total fish consumption, the questionnaires also allowed participants to report consumption of bought and caught fish separately, enabling follow-up actions that focus on local sources of fish.

Conclusions.

Biomonitoring data paired with exposure information can provide valuable insights to guide exposure reduction strategies. Our results indicate that both locally caught and bought fish and shellfish were associated with elevated serum PFAS levels in this study population of API adults in Northern California. Consumption of non-fillet fish parts was strongly associated with serum PFAS levels. Blood mercury was significantly correlated with several serum PFAS levels. These findings suggest a significant route of exposure for a vulnerable population and differences in both consumption and PFAS levels between API sub-populations. Future work should consider additional PFASs that might contribute to PFAS body burdens, other high seafood-consuming populations, and identification of specific locations of caught fish. Continued sampling efforts to expand consumption advisories and outreach efforts to increase awareness of potential health impacts are needed to reduce exposures to contaminants and protect public health.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Data analysis was performed by Kelly Chen. The first draft of the manuscript and visualization of figures was completed by Kelly Chen and Emily Beglarian. Study investigation was performed by Duyen Kauffman, Farmarry Saephan, June-Soo Park, Sabrina Smith, Key-Young Choe, and Jianwen She. Project administration was completed by Duyen Kauffman, Farmarry Saephan, and Kathleen Attfield. Project conceptualization and supervision were performed by Nerissa Wu and Kathleen Attfield. All authors commented on previous versions of the manuscript and approved the final manuscript.

Funding

This study was supported by the Centers for Disease Control and Prevention Cooperative Agreement #5U88EH00148-03 and a one-year state budget augmentation intended for use on environmental justice projects.

Data Availability

The datasets generated during and/or analysed during the current study are not publically available to protect participant privacy but are available from the corresponding author on reasonable request.

Declarations

Ethics Approval and Consent to Participate

The study was approved by the California Committee for the Protection of Human Subjects. Written informed consent was obtained from all study participants.

Conflict of interest

The authors have no relevant financial or non-financial interests to disclose.