Assessment of human exposure to uncommon industrial toxicants in Glynn County, Georgia

Abstract

Coastal Glynn County, Georgia, is home to four hazardous sites on the United States EPA’s National Priorities List. Toxicants of concern include mercury, the pesticide toxaphene, and Aroclor 1268, a mixture of highly chlorinated polychlorinated biphenyls (PCBs); these toxicants are known to persist in the local environment and/or regional aquatic life, including local seafood. At the invitation of, and in partnership with, local community leaders and environmental groups, we conducted a human exposure study in Glynn County. The average age of the study participants was 61 years, 66 % were female, and 46 % were Black. Mercury levels in study participants were comparable to the general US population. Levels of less chlorinated PCBs (PCB 118, 138, 153 and 180) were lower in participants compared to the general population, but the highly chlorinated PCBs associated with Aroclor 1268 were elevated; 19.3 %, 25.0 % and 39.8 % of participants were above the estimated 95th percentile reference values for PCBs 196 + 203, 199, and 206, respectively. About 20 % of participants were above the 95th percentile reference level for both toxaphene Parlars tested (Parlars 26 and 50). We also report on several other toxicants including other metals (lead and cadmium), p,p’-DDE, and poly- and per-fluorinated alkyl substances (PFAS). This study provides evidence that toxicants associated with local hazardous sites have contributed to exposures in Glynn County residents, and that some residents have exposures far exceeding what is common in the general population.

1. Introduction

Glynn County, Georgia, is located on a peninsula roughly midway between Savannah, Georgia and Jacksonville, Florida. The city of Brunswick – the county seat – has a long history of polluting industries and, as a result, is now home to 23 Superfund-regulated sites, of which 17 are active and include three finalized United States Environmental Protection Agency (US EPA) National Priorities Listed (NPL) sites, and one proposed NPL site. (US Environmental Protection Agencya) The four NPL sites are situated in close proximity to each other and close to the coastal rivers and marshes of the city (Fig. 1).

Fig. 1. National Priorities Listed (NPL) sites in Glynn County, GA. The three yellow sites (#1 = Brunswick Wood Preserving, #2 = Hercules 009 Landfill, #3 = LCP Chemicals) are NPL-listed and the red site (#4, Terry Creek) is a proposed site.

Base map downloaded from EPAs Superfund National Priorities List (NPL) Where You Live Map. https://www.epa.gov/superfund/search-superfund-sites-where-you-live#map. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

The three NPL sites of particular concern are the Terry Creek Dredge Spoil Areas/Hercules Outfall, the Hercules 009 Landfill, and the LCP Chemicals site.

Terry Creek is contaminated with toxaphene, an environmentally and biologically persistent pesticide manufactured on-site between the mid-1900s and the 1980s; the Hercules Landfill was used by Hercules LLC for waste disposal and is also contaminated with toxaphene. (Environmental Protection Agencya; Environmental Protection Agencyb) Toxaphene, a mixture of more than 670 chlorinated isomers of camphene and bornanes, was widely applied to cotton crops in the US Southeast, amongst other uses, primarily during the 1970s and 1980s (Agency for Toxic Substances and Disease Registry, 2014a; Agency for Toxic Substances and Disease Registrya; US Environmental Protection Agency, 2000). Toxaphene toxicity is poorly studied in comparison to many other chlorinated pesticides, but like most insecticides, its major effect is through central nervous system overstimulation. (Agency for Toxic Substances and Disease Registrya; US Environmental Protection Agency, 2000; de Geus et al., 1999; US Environmental Protection Agency, 2018) Chronic inhalation exposure in humans has been linked to respiratory toxicity, while chronic oral exposure in animals has resulted in effects on the liver, kidney, spleen, adrenal and thyroid glands, central nervous system and immune system (Agency for Toxic Substances and Disease Registry, 2014a; Agency for Toxic Substances and Disease Registrya). The US EPA classifies toxaphene as a probable human carcinogen (US Environmental Protection Agency, 2000). Toxaphene exposure is not assessed in the National Health and Nutrition Examination Survey (NHANES).

Toxicants of primary concern from the 813-acre LCP Chemicals site are Arcolor 1268 and mercury (Agency for Toxic Substances and Disease Registry, 2014b). Aroclors are mixtures of polychlorinated biphenyls (PCBs), a class of manmade organic chemicals widely used in numerous applications including as dielectric fluids in capacitors and in construction materials such as caulks. (Agency for Toxic Substances and Disease Registryb; Erickson and Kaley, 2011) Production of these chemicals was banned in the United States in 1979, but they remain in our environment (Erickson and Kaley, 2011; Ross, 2004). Arcolor 1268 is composed primarily of highly chlorinated PCBs including PCBs 196 + 203, 199, and 206, with PCB 206 comprising the largest component of the mixture (Kannan et al., 1997; Wirth et al., 2014; Draper et al., 1991; Ishikawa et al., 2005; Maruya and Lee, 1998). PCBs are associated with a range of health effects on the immune, reproductive, nervous, and endocrine systems. (Agency for Toxic Substances and Disease Registryb; Environmental Protection Agencyc; Montano et al., 2022) However, the half-lives and health effects of the highly chlorinated PCBs associated with Aroclor 1268 are not well understood in comparison to some of the less chlorinated PCBs more commonly found in animal-sourced foods (Idowu et al., 2023). As a class, the US EPA considers PCBs as probable human carcinogens (US Environmental Protection Agencyb), but a more recent examination by the International Agency for Research on Cancer classified them as a Group 1 human carcinogen (Lauby-Secretan et al., 2013).

In contrast to toxaphene and Aroclor 1268, mercury is a relatively well-studied toxicant that is naturally occurring. Mercury is neurotoxic and has been linked to a suite of health problems including visual impairment, issues with coordination, and muscle weakness. (Agency for Toxic Substances and Disease Registryc; US Environmental Protection Agencyc; Ha et al., 2017) Mercury exposure in fetuses, infants, and children are of particular concern because of its purported cognitive and developmental impairments (US Environmental Protection Agencyc; Ha et al., 2017). Kidney damage has also been demonstrated. (Agency for Toxic Substances and Disease Registryc; US Environmental Protection Agencyc; Ha et al., 2017)

Evidence of environmental contamination in sediment, soil, water and other environmental media by mercury, Aroclor 1268, and/or toxaphene has been documented in and around Glynn County’s NPL sites (Agency for Toxic Substances and Disease Registry, 2014b; Kannan et al., 1997; EPA, 2017). Coastal wildlife has also shown elevated levels of toxicants, for example in birds and various types of seafood including fish and shellfish (Georgia Department of Natural Resources, 2023; Robinson et al., 2015; Kannan et al., 1998; Maruya et al., 2001). As a result, Georgia’s Fish Consumption Guidelines recommend restricting intake of many of Brunswick’s locally caught seafood species. Local populations of dolphins, which feed on Brunswick-area fish, have some of the highest levels of PCBs reported (Balmer et al., 2011; Backer et al., 2019); seafood consumption is widely assumed as one key pathway for human exposure in the area.

Despite Brunswick’s history of environmental contamination, to our knowledge no human biomonitoring study has been conducted. A study on the health effects of occupational mercury exposure in LCP workers reported mixed results (Frumkin et al., 2001), and another study of mercury exposure in local fishers is considered to be flawed for its lack of representativeness and methodological concerns including its reliance on urine, potential for recall bias, and the choice of reference level (Agency for Toxic Substances and Disease Registry, 2014b). One small study (n = 9) of residents on nearby Sapelo Island reported median levels of several of the PCBs associated with Arcolor 1268 above the 95th percentile of the US population (Backer et al., 2019). That study also documented high levels of certain per- and polyfluoroalkyl substances (PFAS) in the Sapelo residents.

Here we report the results of a community-engaged human exposure study of Glynn County residents initiated at the request of local community leaders and environmental groups. We focus on mercury, Aroclor 1268, and toxaphene, but also report on a suite of other toxicants that either were easily analyzed alongside those target compounds (other metals, PCBs and pesticides) or are an emerging concern (PFAS).

2. Methods

2.1. Study population

We enrolled 100 Glynn County, GA residents to participate in our cross-sectional exposure study in 2023. Participants were eligible for inclusion if they were ≥18 years of age, were current residents of Glynn County, and lived in Glynn County for ≥10 years. The participants were recruited through word-of-mouth and via reports from local media that publicized the study. Of those who expressed interest in the study, we prioritized enrolling those living in neighborhoods adjacent to one of the three NPL sites, those who reported current fishing activities, and those who worked at one of the sites or lived with someone who did. Two participants were enrolled who did not meet the inclusion criteria, one who lived in Glynn County for less than 10 years and one who worked but did not live there; we nonetheless included them in our analysis to maximize statistical power. This study was approved by the Emory University Institutional Review Board (STUDY00005384).

After obtaining written, informed consent, participants completed one in-person study visit in March 2023. At the study visit, participants completed a short questionnaire that collected sociodemographic information and potential pathways of exposure (e.g., working or living with someone who worked at the sites, residential histories, fishing practices). Each participant provided two venous blood samples via venipuncture by a trained phlebotomist into a 2 mL, blue-top certified metal-free whole blood vacutainer (collected first) and a 10 mL tiger-top serum separator vacutainer. All vacutainers were polyethylene terephthalate plastic. Samples were refrigerated and then hand carried on ice to Emory University. At Emory, the serum separator tube was kept at room temperature for 30 min and allowed to clot. The tube was then spun down in a centrifuge at 10,000 rpm and the serum layer was transferred to a clean vial. All serum samples were then stored at −20 °C in glass vials, and all whole blood samples were kept refrigerated until analysis in the Laboratory of Exposure Assessment and Development in Environmental Research at Emory University. A subset of participants (n = 21) did not provide enough blood sample to test for all three target toxicants (mercury, Aroclor 1268, and toxaphene) and therefore were offered a second blood draw in January 2024; nine accepted. In total, 96 participants were tested for metals, 88 for PCBs/pesticides, and 34 for PFAS.

2.2. Analysis of metals

Lead (Pb), cadmium (Cd), and total mercury (Hg) were quantified in whole blood using inductively coupled plasma-mass spectrometry (ICP-MS) (Agilent 7700, Agilent Technologies, CA, USA). A 250 μL aliquot of each sample was microwave digested with ultra-purified nitric acid before dilution with 2 % ultra-purified nitric acid and a diluent solution containing a mixture of the following internal standards (ISTD): rhodium, indium, lutetium, and iridium. The final volume of sample before injection was 10 mL. The digested samples were then analyzed via ICP-MS alongside a matrix-based calibration curve (ranging from 0.05 to 50 ng/mL for Pb and Cd, and 0.05–2.5 ng/mL for Hg), 3 analytical blanks, 2 matrix blanks, and 2 quality control samples. The quality control samples consisted of Milli-Q water (Merck Millipore, Darmstadt, Germany) spiked at 2 levels. To ensure the accuracy of the data produced, certified reference materials obtained from the National Institute of Standards and Technology (NIST) (Gathersburg, Maryland, USA) (SRM 955c Toxic Elements in Caprine Blood) and RECIPE Chemicals (München, Germany) (ClinChek^® Whole Blood Control for Trace Elements) and were analyzed alongside the samples. The observed concentrations of the target metals in these reference materials were within ±20 % of the certified concentrations. During the analysis, polyatomic spectral interferences were removed using a dynamic reaction cell. Concentrations of the target metals were derived from an equation defining the relative response of native analytes to the response to the ISTDs across the calibrant concentration range. These concentrations were later converted to final concentrations (ng/mL) after accounting for the dilution factor and background concentrations.

The method limits of detection (LODs) are provided in Supplementary Table 1 and ranged from 0.139 ng/mL for mercury to 0.518 ng/mL for cadmium. The method precisions, calculated from values of replicate QC samples and expressed as relative standard deviations (RSD), were below 10 %.

2.3. Analysis of PCBs, toxaphene, and p,p’-DDE

Seven PCB congeners, two toxaphene pesticide congeners (Parlars 26 and 50) and p’p’- dichlorodiphenyldichloroethylene (p,p’-DDE) were analyzed in serum samples using previously published methods (Barr et al., 2004; Chen et al., 2014; Marder et al., 2016). Detailed information on these methods can be found in the original publications. Amongst the PCBs, we included some that are more ubiquitous in the US population (PCB 118, 138, 153, 180) as well as several more specifically associated with Aroclor 1268 (PCB 196 + 203, 199 and 206). The method used to measure PCBs and p,p’-DDE was modified slightly to include these additional Aroclor 1268-related PCB congeners. Method information about the additional four congeners is provided in Supplementary Table 2.

Briefly, 1 mL of serum was mixed with isotopically labeled analogs of the target chemicals. After adding a sodium sulfate/1-propanol solution, the sample was loaded onto a silica-based C18 solid-phase extraction (SPE) cartridge that was pre-conditioned. The cartridge was washed, and the target analytes were eluted with an ethyl acetate/n-hexane solution. A residual aqueous layer was discarded, and anhydrous sodium sulfate was added to remove the remaining water. The eluate was loaded onto a Florisil^® SPE cartridge and eluted with an ethyl acetate/n-hexane solution to remove residual biogenic material from the extract. The extract was concentrated to dryness and reconstituted in nonane prior to injection analysis using gas chromatography-tandem mass spectrometry (GC-MS/MS) (Agilent Technologies, Santa Clara, CA). For separation of the target toxicants, a Phenomenex ZB-5 analytical column (30 m × 0.25 mm × 0.25 μm) was used (Phenomenex, Torrance, CA). PCBs/p,p’-DDE and toxaphene congeners were analyzed separately, using a subsequent injection of 2 μL in the pulsed splitless mode. PCBs/p,p’-DDE were analyzed using electron impact ionization in the multiple reaction monitoring mode (MRM). The toxaphene congeners were analyzed using negative chemical ionization (NCI) coupled with selected ion monitoring. High-purity helium was used as the carrier gas for both analyses. Methane was used as the NCI reagent gas. For the MRM analysis, ultra purified nitrogen (>99.99 %) was used as a collision gas.

During mass spectrometric analysis, one quantitation and one confirmation precursor to product ion transition were monitored for the native analytes. One quantitation precursor to product ion transition was monitored for the labeled analog. Concentrations of the target analytes were quantified using isotope dilution calibration. The LODs were 20 pg/g for p,p’-DDE, 5 pg/mL for the two toxaphene Parlars, and 19–40 pg/mL for the PCB congeners (Supplementary Table 1).

In each analytical run, participant samples were analyzed concurrently with a 10-point external solvent-based calibration curve, one solvent blank sample, one matrix “blank” sample consisting of unspiked serum which also naturally contained low levels of PCBs 118, 138, 153 and 180 and p,p’-DDE, and two matrix-based QC (spiked serum) samples (2 levels). The method showed good precision as measured in both QC and NIST SRM samples (<15 % relative standard deviation (RSD)). The analysis of NIST SRM 1958 replicates indicated that the accuracy of the method was within 20 % of the certified values. For p,p’-DDE, interlaboratory reproducibility was evaluated and certified by successful participation in the German External Quality Assessment Scheme for Analyses in Biological Materials (G-EQUAS).

2.4. Analysis of PFAS

PFAS were analyzed using an online solid phase extraction followed by liquid chromatography-tandem mass spectrometry analysis (LC-MS/MS) (Honda et al., 2018; Liang et al., 2023; Chang et al., 2021). A 100 μL aliquot of each sample was mixed with an isotopically labeled internal standard mixture in methanol (Cambridge Isotope Laboratories (Andover, MA)) for protein precipitation and separation. After centrifugation of the protein precipitated serum, 240 μL of the supernatant was mixed with 360 μL of 0.1 M formic acid in a sample vial for a total sample volume of 600 μL. Each diluted sample (400 μL) was injected onto a Strata RP (Phenomenex, Torrance, CA) on-line extraction column (2.1 × 20 mm). The on-line extraction column was washed with 0.1 M formic acid: acetonitrile (90:10, V/V), then the target analytes were transferred from the on-line extraction column to a Betasil C18 (ThermoFisher, Waltham, MA) analytical column (4.6 × 150 mm, 5um) for chromatographic separation and analysis by LC-MS/MS (Agilent 6460, Agilent Technologies, Santa Clara, CA) in the MRM mode.

Quantification of the target analytes was performed using isotope dilution calibration. The target analytes were: linear and branched perfluorooctanesulfonic acid (PFOS), perfluorooctanoic acid (PFOA), perfluoroundecanoic acid (PFUnDA), perfluorononanoic acid (PFNA), perfluorohexanesulfonic acid (PFHxS), and perfluorodecanoic acid (PFDA). In each analytical batch, a matrix-based calibration curve (ranging from 0.01 to 20 ng/sample volume), a solvent blank sample, a matrix blank sample, and QC samples (pooled serum spiked at 2 levels) were prepared and injected alongside the participant samples. The LODs ranged from 0.007 to 0.1 ng/mL (Supplementary Table 1). The method RSDs were below 15 % based upon the QC materials. The analysis of NIST SRM 1958 replicates indicated that the accuracy of the method was within ±20 % of the certified values. Inter-laboratory reproducibility was evaluated and certified by successful participation in the G-EQUAS.

2.5. Statistical analysis

LODs were calculated as the lowest standard measurable with a signal-to-noise ratio of 3. When a discernible blank level was present, the LOD was set at the blank concentration plus 3s₀. For each toxicant, values below the LOD were imputed using LOD/√2. Levels were not adjusted for blood lipids, as this approach assumes all participants had the same baseline blood lipid value, which is unlikely, and as such has experienced criticism for its potential bias (Schisterman et al., 2005).

We examined the distribution of demographic characteristics in our study population using means, standard deviations (SDs), frequencies, and counts. Because our data were not normally distributed, we converted them to their natural logarithm before further statistical analyses. We then assessed the distribution of toxicants using geometric means (GM) and geometric standard deviations (GSDs). We calculated Spearman correlation coefficients to estimate correlations between individual toxicants.

We then assessed whether any participants in our study had toxicant levels that were above the estimated 95th percentile of the general population. In this analysis, reference values for the general population were as follows.

• Reference levels for metals and PFAS were the geometric mean and 95th percentile levels from the 2017–2018 cycle of the National Health and Nutrition Examination Survey (NHANES) for all adults aged 20+. (US Centers for Disease Control and Prevention)

• Reference levels for PCBs and p,p’-DDE were last measured in individuals in the 2003–2004 NHANES cycle. We derived the geometric mean and 95th percentile levels of adults aged 40+, which we calculated from the raw data (www.cdc.gov/nhanes data files DEMO_C, L28DFP_C, L28OCP_C, and L28NPB_C) and then adjusted by assuming that levels have declined by one half-life since that time; to our knowledge there are few (if any) human half-life estimates in the literature for the highly chlorinated PCBs, but we note that estimates for the less chlorinated congeners tend to be < 20 years (Idowu et al., 2023).

• As no population-based reference levels are available for toxaphene in the United States, we used reference levels of Parlars 26 and 50 from the 2007–2009 national survey from Health Canada for adults aged 60–79 (Health Canada, 2010)—the only age for which the reported values were above the LOD—again with the assumption that levels have declined by one half-life since that time.

Reference levels before and after adjustments are reported in Supplementary Table 3.

Logistic regression was used to assess relationships between sociodemographic characteristics and exposure to the a priori toxicants of concern. The sociodemographic characteristics included as exposures in these analyses were race, age, sex, and income; each association is reported after adjusting for the others. The potential pathways of exposure assessed were water source, local seafood consumption, current fishing, whether the person ever worked at an industry operating on an NPL site or lived with someone who did, and living in proximity to the NPL sites; each pathway is reported after adjusting for race, age, sex, and income. Each outcome (i.e., individual toxicant of concern) was coded as binary (above or below the limit of detection for each toxicant) and modeled separately. For ease of interpretation, beta estimates obtained from logistic regression models were exponentiated to obtain odds ratios which can be interpreted as the odds of participant exposure being above the limit of detection. In supplementary analyses, we conducted analogous analyses using linear regression which included continuous toxicant concentrations as the outcome, which can be interpreted as the average change in exposure level given a unit change in a given characteristic or exposure pathway.

For the analyses which included living in close proximity to an NPL site as the exposure, present and past addresses were derived from the questionnaire and geocoded in R using the geocoding service from ArcGIS and package “tidygeoder.” Street address, city, county, state, and zipcode were geocoded into latitude and longitude values. We excluded addresses with latitude and longitude values outside of Georgia, indicating erroneous recording.

3. Results

The average age of participants included in the study was 60.8 years (SD = 14.7; Table 1). Two-thirds of participants were female, just over half were White Non-Hispanic, and the average time lived in Glynn County was 46.7 years. Twenty-four percent of participants reported current fishing activities. Of the 97 participants that provided an adequate blood sample, 7 % reported working at one of the NPL sites and 26 % reported living with someone who did (Table 1).

Table 1.

Descriptive statistics of study participants (n=97).

Variable	Mean (SD) or N (%)
Age in years	60.8 (14.7)
Years resident in Glynn County	46.7 (20.2)
Female sex	64 (66.0 %)
Male sex	33 (34.0 %)
Black/Hispanic/Other racea	45 (46.4 %)
White, Non-Hispanic race	52 (53.6 %)
Household income ≥ $50K/year	53 (54.6 %)
Use well water	23 (23.7 %)
Currently fish	23 (23.7 %)
People who eat local seafood ≥ once/week	29 (29.9 %)
People who worked at an industry operating on an NPL site	7 (7.2 %)
People who lived with a worker at an NPL site	25 (25.8 %)

^aOf these participants, all but five self-reported as Black/African-American.

SD = standard deviation.

The GMs and GSDs of the participants’ toxicant levels are reported in Table 2. Participant blood levels of metals, including mercury, were similar to the US population (Table 2).

Table 2. Geometric means and geometric standard deviations (SD) of toxicant blood levels in the study population and reference populations.

96 participants were tested for metals, 88 for PCBs and pesticides, and 34 for PFAS.

	Number (%) of participants > LOD	Reference geometric meana	Study geometric mean (SD)	Reference 95th percentilea	Number (%) of participants >95th percentile
Metals (μg/dL)	N = 96 (100 %)
Lead	96 (100 %)	0.9	0.9 (1.9)	2.6	5 (5.2 %)
Mercury	95 (99.0 %)	0.8	0.8 (2.5)	4.4	1 (1.0 %)
Cadmium	56 (58.3 %)	0.3	0.5 (2.0)	1.4	9 (9.4 %)
PCBs (pg/g)	N = 88 (100 %)
PCB 118	29 (33.0 %)	36.1	21.8 (1.9)	180.2	0 (0 %)
PCB 138	73 (83.0 %)	101.7	51.2 (2.1)	398.3	0 (0 %)
PCB 153	75 (85.2 %)	139.4	61.5 (2.2)	522.1	0 (0 %)
PCB 180	70 (79.5 %)	123.5	62.3 (2.6)	396.6	3 (3.4.%)
PCB 196 + 203	24 (27.3)	22.3	45.5 (2.0)	73.9	17 (19.3 %)
PCB 199	49 (55.7 %)	28.3	43.4 (3.2)	101.9	22 (25.0 %)
PCB 206	62 (70.5 %)	17.8	52.9 (2.8)	71.0	35 (39.8 %)
Pesticides (pg/mL) c	N = 88 (100 %)
Toxaphene 26d	19 (21.6 %)	5.0	5.7 (2.1)	5.0	19 (21.6 %)
Toxaphene 50	31 (35.2 %)	5.0	6.2 (2.0)	10.0	18 (20.5 %)
p,p’-DDE	88 (100 %)	1808.5	462.2 (3.1)	12773.0	1 (1.1 %)
PFAS (ng/mL)	N = 34 (100 %)
PFOS (linear + branched)	34 (100 %)	4.5	5.0 (2.6)	15.1	3 (8.8 %)
PFOA	33 (97.1 %)	1.5	0.7 (2.2)	3.9	1 (2.9 %)
PFNA	32 (94.1 %)	0.4	0.5 (2.5)	1.4	3 (8.8 %)
PFUnDA	23 (67.6 %)	0.1	0.2 (2.2)	0.4	4 (11.8 %)
PFHxS	34 (100 %)	1.1	0.7 (2.1)	3.8	0 (0 %)
PFDA	34 (100 %)	0.2	0.4 (2.5)	0.6	11 (32.4 %)
Total PFASb	–	7.8	8.0 (2.3)	25.2	3 (8.8 %)

^aReference values for metals and PFAS are from the 2017/18 NHANES; reference values for PCBs and p,p’-DDE are from 2003/2004 NHANES, but adjusted for time since survey; the reference value for toxaphene is from a 2007–2009 survey from Health Canada, but adjusted for time since survey (see Methods for details).

^bDefined as the sum of the six individual PFAS reported in the table.

^cRaw lab data for p,p’-DDE was reported as pg/g whereas toxaphene was pg/mL. For consistency, we report p,p’-DDE also as pg/mL, noting the two are highly similar but not identical (1 mL serum = 1.026g serum).

^dA subset of participants (n = 11) whose blood sample was small had imputed levels >5; they are not included as having levels above the 95th percentile, though it does affect the interpretation of the geometric mean (see Supplementary Table 1 for LODs).

For PCBs, participants generally had relatively low levels of the more ubiquitous congeners (PCBs 118, 138, 153, 180), but had elevated levels of the highly chlorinated congeners associated with Aroclor 1268 (PCBs 196 + 203, 199 and 206). The mean level of PCB 206 – the primary constituent of Aroclor 1268 – was approximately triple the estimated reference mean, with nearly 40 % of participants above the estimated reference 95th percentile. There was substantial variation amongst these highly exposed participants, with several individuals recording levels many times higher than the 95th percentile reference values (Fig. 2).

Fig. 2. Blood levels of target toxicants (mercury, PCBs and toxaphene).

Blue dots indicate values above the geometric mean of the reference population while red are above the 95th percentile. Note that the reported population geometric mean and 95th percentile reference values for toxaphene Parlar 26 were the same (5 pg/mL), and so all the blue dots could also be considered red. Reference values for mercury are from the 2017/18 NHANES; reference values for PCBs are from 2003/2004 NHANES, adjusted for time since survey; the reference value for toxaphene is from a 2007–2009 survey from Health Canada, adjusted for time since survey (see Section 2.5 of the Methods for details). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Mean toxaphene levels in the study participants were also above reference means, with a high percentage (21.6 % for Parlar 26 and 20.5 % for Parlar 50) above the 95th percentile values (Table 2, Fig. 2). p,p’-DDE levels were generally below US population levels.

Spearman correlations were uniformly positive among the PCBs (range: 0.25–0.95) (Supplementary Fig. 1). The three highly chlorinated PCBs were highly correlated with each other (0.68–0.89) – particularly PCB 206 and PCB 199 (0.89) – indicating a common source (Aroclor 1268). Some of the less chlorinated PCBs also had high correlations with p,p’-DDE. The PCBs were not strongly correlated to metals or the toxaphene Parlars.

For individual PFAS, PFDA was detected in all participants tested and was the type most elevated in our study population, with the study geometric mean twice the reference mean and with 32.4 % of participants above the reference 95th percentile.

For all three of the highly chlorinated PCBs analyzed, older participants were more likely to have detectable levels than younger participants (all p < 0.01) (Fig. 3, Supplementary Table 4). Black/Hispanic/Other participants were more likely to have detectable levels than White Non-Hispanic participants for PCBs 199 (OR = 3.81, 95 %CI 1.21–14.8) and 206 (OR = 4.30, 95 %CI 1.28–17.46) with suggestive evidence for PCB 196 + 203 (OR = 2.39, 95%CI 0.81–7.45). When this analysis is conducted on White vs Black participants, results are similar (Supplementary Table 5). No groups had clearly elevated odds of detection for the two toxaphene Parlars. Results from the linear regressions show qualitatively similar results (Supplementary Fig. 2).

Fig. 3. Odds ratios (OR) and 95 % confidence intervals for the likelihood that select sociodemographic groups had exposure levels above the limit of detection compared to a corresponding reference (Black/Hispanic/Other participants are compared with White Non-Hispanic participants, males with females, age >60 with age ≤ 60, and those with >50K income with ≤ 50K).

Results are from logistic regression models with the result of each sub-group controlled for the others; for example, results in participants age >60 vs age ≤60 are controlled for race, sex, and income. The dashed horizontal line is where the OR = 1. The same results are reported in table form in Supplementary Table 4.

When evaluating the influence of potential exposure pathways (Fig. 4, Supplementary Table 4), we found that the odds of a detection were significantly higher for PCBs 199 (OR = 5.25, 95 %CI 1.21–28.54) and 206 (OR = 8.52, 95 %CI 1.65–62.18) in people who reported currently fishing compared to those who did not. Self-reported consumption of local seafood did not show a clear trend. Participants with city water as their primary drinking water source generally had a lower, albeit non-significant (p > 0.05), odds of detection compared to those who used well water. Participants who worked at LCP chemicals (where Aroclor 1268 was used), or lived with someone who did, had a much higher probability of detection for the highly chlorinated PCBs compared to those who did not work or live with someone who worked at LCP chemicals, though confidence intervals were wide due to a small sample size. Linear regression results are reported in Supplementary Fig. 3.

Fig. 4. Odds ratios (OR) and 95 % confidence intervals for the likelihood of exposure levels above the limit of detection by potential pathway of exposure for sub-groups compared to the corresponding reference group.

The dashed horizontal line is where the OR = 1. The same results are reported in table form in Supplementary Table 4.

There were no clear associations with residential proximity to the hazardous sites (Fig. 5).

Fig. 5. Odds ratios (OR) and 95 % confidence intervals for the likelihood of exposure levels above the limit of detection in study participants who ever lived <2 miles from a given NPL site.

The dashed horizontal line is where the OR = 1.

4. Discussion

In this human exposure study of residents of Glynn County, Georgia, we found evidence of elevated blood levels of two classes of toxicants with known local sources. One class was the highly chlorinated PCB congeners comprising Aroclor 1268, a PCB mixture used at the LCP Chemicals NPL site. The mean level in study participants was higher than the estimated reference level for all three of the highly chlorinated congeners we investigated (PCB 196 + 203, 199 and 206), but was especially notable for PCB 206, which is a dominant component of Aroclor 1268 (Kannan et al., 1997; Wirth et al., 2014; Draper et al., 1991; Ishikawa et al., 2005; Maruya and Lee, 1998). The less chlorinated congeners that are more commonly derived from food chain consumption of dairy and meat were similar or lower than population-based levels, suggesting a separate source of exposure for the Aroclor 1268 PCB congeners. The other toxicant showing indications of elevated levels was the pesticide toxaphene, which was produced at the Terry Creek NPL site, and which also contaminated the Hercules Landfill NPL site. To our knowledge, the NPL sites were the only meaningful point sources of these toxicants locally.

A key motivation behind community requests for this study was environmental justice concerns, as the three NPL sites are located primarily in low-income, predominantly Black neighborhoods. We note that race is a complex construct that often reflects broader societal inequalities. We found indications that blood levels of the highly chlorinated PCBs were indeed higher in Black versus white participants, even after controlling for age, sex, and income. Possible reasons for higher levels in Black participants in Glynn County may include residential patterns (discussed below) as well as fishing or seafood consumption practices, which are strongly tied to Gullah/Geechee culture and heritage.

Unlike the other a priori toxicants of concern, we did not find evidence of elevated mercury levels. Blood analysis of mercury only reflects relatively recent (past few months) exposure (Yaginuma-Sakurai et al., 2012) as compared to PCBs and toxaphene, where blood levels estimate long-term (years to decades) exposures; this may indicate that exposures from the NPL sites are largely historical. However, other explanations are also possible; for example, it may be that ongoing remediation efforts are more successful for mercury than the other toxicants, or that fate and transport and/or pathways of exposure are (or were) different.

Overall, our findings for PCBs and mercury are consistent with what was reported in the small study of nine residents of nearby Sapelo Island, which also found elevated levels of Aroclor 1268-related PCBs, but not mercury (Backer et al., 2019). That study specifically enrolled high seafood-consuming participants but did not include measurement of toxaphene. That study also reported high levels of PFDA, which we also found in terms of a high rate (100 %) of detection and a higher geometric mean compared to NHANES; results were less consistent for other types of PFAS.

The finding of more highly chlorinated PCB and toxaphene levels in older participants is not surprising, as these are persistent chemicals that bioaccumulate and biomagnify over time. Older individuals have a longer period to be exposed and were alive when these chemicals were still in use.

Potential pathways of exposure identified a priori included fishing/consumption of local seafood, occupational exposures, and residential proximity to the NPL sites. Participants who fished had higher average blood levels for several of the focal toxicants. Why fishers had higher levels, but people who reported eating more local seafood did not, is an important question for future work, but may relate to problems with recall, difficulties knowing where exactly seafood comes from, differences in species consumed, or it may be that store-bought local seafood in Brunswick differs from what is self-caught.

Our finding of higher blood levels of the PCBs associated with Aroclor 1268 in people who worked at LCP Chemicals, or lived with someone who did, is not surprising. People working at an industry that uses hazardous substances may be occupationally exposed through their day-to-day activities, and conversations with study participants (and other community members) revealed that family members often came home from work smelling of chemicals. In some cases, they described workers bringing products home for household use (e.g., toxaphene to treat lawns for pests, or liquid mercury for kids to play with).

There was no clear association between residential location and toxicant blood levels. This also warrants further investigation, for example with more study participants and with analyses that account for the length of time lived in close proximity to the contaminated sites. Considering that the polluting industrial processes have not been active for multiple decades, it may be that historical residences may be the most relevant. It is also possible that other factors are more important than residential locations, for example dietary habits (e.g., seafood consumption), or if contaminants are being redistributed around the county in environmental media (e.g., surface waters or floodwaters).

Our study should be interpreted in light of its limitations. One key limitation was the modest sample size, which contributed to uncertainty in some of the sub-group analyses; a larger study is needed to enable a more targeted investigation of specific groups (e.g., fishers or those occupationally exposed). There were also limitations in the questionnaire that served as the basis of the sub-group analyses, which was produced for this study and was not explicitly validated. For instance, we observed that participants faced difficulties recalling residential histories and food consumption habits, so recall visuals or more structured, validated questionnaires may have yielded more conclusive results. Most questions, however, were simple and straightforward, for example, those asking participant age, sex and race, which limits recall bias.

Another key challenge was the lack of recent reference data. Peer-reviewed reports of blood levels of Aroclor 1268 congeners and toxaphene are generally old and/or only for specific populations. For example, one study from the US (Barr et al., 2004) reported toxaphene levels in a few US cities based on samples collected in the 1980s and 1990s, while another from Japan reported levels in pregnant women collected in the early 2000s (Araki et al., 2018). A study from a PCB-polluted region of eastern Slovakia reported results for some highly chlorinated PCBs, but not PCB 206 (Strémy et al., 2019). In contrast, a 1996 study of middle-aged men in southern Norway reported levels for PCB 206, but not 199 or 196 + 203 (Johansen et al., 1996).

As a result, for reference levels of PCBs and toxaphene, we relied on pre-2010 surveys from the USA and Canada, respectively, that we adjusted by assuming a reduction in exposure by one half-life. Although this is a somewhat crude approach, it is easy to understand and comparable to some other possible assumptions, for example using a 4 % annual decline for persistent organic pollutants, which was recently estimated from Norwegian data (Nøst et al., 2019). To our knowledge, there are few, if any, human half-life estimates in the literature for the highly chlorinated PCBs, but we note that estimates, at least for the less chlorinated congeners, tend to be < 20 years (Idowu et al., 2023). We do not know of any human half-life estimates for toxaphene. Different reference values could change the quantitative interpretation of the results, but we posit that for any plausible population-based 95th percentile reference value, the highest exposed individuals in our study population would likely still be far above that level.

Many future directions exist that could build upon this work. A more expansive study would be valuable from a purely exposure standpoint, but could also be accompanied by epidemiological studies to link the exposures to health outcomes. Epidemiological studies of highly chlorinated PCBs and toxaphene are rare and critically needed. Environmental sampling would also be beneficial to triangulate potential pathways of exposure and, importantly, to understand whether exposure is ongoing; enrolling younger participants would also help.

Key strengths of the study include the deep collaboration between community members and the research team, the recruitment of a diverse study population, and a focus on uncommon but highly persistent toxicants, some of which are not routinely monitored in national surveys and required the development of specialized laboratory methods to analyze.

Supplementary Material

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envpol.2025.126240.

Data availability statement

A de-identified dataset can be made available for research purposes upon reasonable request to the corresponding author, Dr. Noah Scovronick, at scovronick@emory.edu. All data requests are subject to review by the Emory University Institutional Review Board (IRB).