PERSONAL AND ENVIRONMENTAL PREDICTORS OF POLYCYCLIC AROMATIC HYDROCARBON EXPOSURE IDENTIFIED THROUGH REPEATED SILICONE WRISTBAND SAMPLING

Abstract

This study integrates quantitative data on personal exposure to polycyclic aromatic hydrocarbons (PAHs) in 162 silicone wristbands with demographics, behavioral information, and housing characteristics to explore contributions to residential exposure in a community influenced by historic and current industrial activities over the course of a year. Forty-six residents completed questionnaires and wore silicone wristbands as personal passive samplers for seven consecutive days on up to four separate occasions between November 2022 and June 2023; the repeated measures in this study lend insight into exposure sources with variability. It was hypothesized that individual behaviors and housing characteristics would be sources of dependence and correlation between personal PAH exposures. Fifty PAHs were detected, seventeen of which were alkylated PAHs. Exposure to PAHs of similar molecular weight was often correlated, notably between naphthalenes (2-rings) and PAHs of 3 or more rings. Generalized linear mixed models identified flooring type, participant age, and sampling month as predictors of increased PAH exposure. Flooring type and use of wood stoves or heavy machinery were identified as predictors of increased naphthalene exposure relative to larger PAHs. Personal behaviors and housing characteristics were sources of dependence and correlation between personal PAH exposures across repeated measures. We demonstrate the value of collection and integration of questionnaire data with exposure data, and repeated measures that consider intra-individual variability. Identification of influential exposure factors through repeated measures of chemical exposure and characterization of variability in personal exposure as performed in this study is important in the development of exposure mitigation strategies.

Graphical Abstract

1. Introduction

Polycyclic aromatic hydrocarbons (PAHs) are a class of semi-volatile organic compounds with multiple fused aromatic rings. PAHs are pervasive, mobile, and persistent pollutants distributed in water, soil, and air. Chemicals in this class can be petrogenic or pyrogenic, including both natural (e.g. wildfires, crude oil) and anthropogenic (e.g. industrial activities, vehicle exhaust) sources (Kim et al., 2013; Patel et al., 2020). Personal ambient PAH exposure depends on outdoor and indoor environments; PAH sources in indoor air are important to account for in personal exposures as North Americans tend to spend an estimated 90% or more of their time indoors, and PAH concentrations are commonly measured to be two to four times higher in indoor than outdoor air (Ma and Harrad, 2015; Rivera et al., 2025; U.S. Environmental Protection Agency, 1989). Major indoor sources of PAHs include smoking, residential heating, cooking, and burning candles or incense (Krugly et al., 2014; Ma and Harrad, 2015). A review by Račić et al., (2025) noted that PAH concentrations in residential indoor environments show considerable variability, driven by regional differences, building characteristics, and indoor activities (Račić et al., 2025).

Health risks associated with PAHs range from acute (e.g. membrane and skin) irritation to chronic toxicity; some are well known carcinogens, mutagens, or teratogens (Abdel-Shafy and Mansour, 2016; Kim et al., 2013). An important toxicological consideration for the bioavailability and exposure routes of atmospheric PAHs is that they can be bound to particulate matter or exist in the more bioavailable vapor phase, and the ratio of bound-to-unbound PAHs depends on environmental conditions and chemical properties (Abdel-Shafy and Mansour, 2016; Kim et al., 2013; Paulik et al., 2018). Warmer temperatures lend to a greater proportion of PAHs in the vapor-phase, with smaller PAHs (i.e. higher vapor pressure) prone to the greatest vaporization (Harrison et al., 1996; Krugly et al., 2014). Currently, insufficient data is available on individual human exposure to vapor-phase PAHs across seasonal conditions. Further, much of the exposure data that exists is limited to a short list of PAH analytes, typically informed by the US EPA’s list of “16 Priority PAHs,” which omits toxicologically relevant alkylated and heterocyclic PAHs (Andersson and Achten, 2015; Mueller et al., 2019).

Silicone wristbands have been used as personal passive sampling devices in vapor-phase exposure monitoring scenarios since first established by O’Connell et al., in 2014 (O’Connell et al., 2014; Samon et al., 2022). Personal monitoring with the silicone wristbands is a useful tool for understanding individuals’ contaminant exposures and giving more accurate exposure estimates than questionnaires or stationary monitoring alone (Donald et al., 2019; Paulik et al., 2018). The combination of silicone wristband samplers and questionnaire data allows for investigation of individual exposures in the context of information about individuals’ behaviors and environments (McLarnan et al., 2024).

The aim of this study is to explore the influence of demographics, indoor environment, and personal habits on total vapor-phase PAH exposure. We assessed a large library of parent and alkylated PAHs in silicone wristbands from 46 residents in St. Helens, OR over eight months to better understand contributions to, and changes in personal exposure. We hypothesized that individual behaviors and housing characteristics are sources of dependence and correlation between personal PAH exposures across repeated measures. While this study focused on integrating questionnaire variables with repeated chemical measures, inter- and intra-individual exposure variability for this dataset and the value of repeated measures in study design is fully detailed in Bramer et al (2025). Determining factors contributing to personal PAH exposure is not only relevant to the residents in St. Helens, OR, but could indicate more general common contributions to personal PAH exposure, seeing as the variables captured in the questionnaire are not unique to this community. The magnitude and types of PAH analytes, including sixty-two parent and alkylated PAHs, and 162 personal sampler wristband samples over eight months of repeated measures is valuable, and lends insight into exposure sources and individual exposure variability.

2. Materials and Methods

A schematic of the study design, data collection, and analysis methods is provided in Figure S1.

2.1. Site Description

The port city of St. Helens, Oregon (population 14,095) is located 86 river miles from the Pacific Ocean, bordering the Multnomah Channel on the Willamette River near the confluence with the Columbia River. Highway 30 and a railroad freight line intersect the city, which contribute to the city being an important transport center. These transport lines, historic industrial activities (City of St. Helens, 2024; Port of Columbia County, 2025), as well as other current natural and anthropogenic sources of pollution are all factors suspected to influence total vapor-phase PAHs in outdoor ambient air in the St. Helens community and the exposure for residents, in addition to personal behaviors and sources in other environments (e.g., work and home).

Across sampling months, average air temperatures ranged from 4.04 ± 1.44°C (February 2023) to 22.5 ± 1.03 °C (May 2023) with winds predominately from the SE (Figure S2 and S3); precipitation, fog, and air quality trends were also captured in Figure S3. Environmental conditions are summarized by sampling period in Tables S1 and S2.

2.3. Participant Recruitment and Communications

All research activities were approved by the Institutional Review Board at Oregon State University (IRB-2020-0529). All volunteer participants provided written or verbal consent prior to data collection. Communication throughout the study was maintained through email and mail with participants and updates to the publicly available study website. Data report-backs were sent to participants with individual and summary data.

2.2. Sampling design

Participating residential addresses ranged from 0.3 to 4.3 km from a central point in the city of St. Helens. Residents of St. Helens wore silicone wristbands as personal passive samplers for seven consecutive days at a time. Participants were split into two sampling groups with alternating sampling months. In total, each participant was asked to wear a wristband during four alternating months, and the two groups together covered eight months of sampling from November 2022 to June 2023 (Table 1).

Table 1.

Sampling scheme and summary of samples analyzed.

Sampling Period	1	2	3	4	5	6	7	8
Dates	11/10-18	12/5-12	1/12-19	2/10-17	3/10-17	4/7-14	5/12-19	6/9-16
	2022	2022	2023	2023	2023	2023	2023	2023
Participant group	1	2	1	2	1	2	1	2
Participant wristbands (n)	18	18	22	18	24	16	24	22
Households that contributed samples	13	12	16	13	17	12	17	15

All silicone wristband samples were mailed to participants and mailed back to the OSU lab with pre-paid shipping at the end of the sampling period. Samples were sealed in airtight polytetrafluoroethylene bags (Welch Fluorocarbon, Dover, NH, USA) for mailing at ambient temperatures; samples were stored at −20 °C in the same containers once returned to the lab, prior to analysis (Anderson et al., 2017; O’Connell et al., 2014). In total, 162 silicone wristband samples were collected and analyzed for PAHs.

Each participant also responded to a one-time baseline questionnaire with 35 questions about their demographics, occupation, environment, and habits that may influence exposure to PAHs before sampling (Figure S1, questionnaire included in full as Supplemental File). With each sample, participants were asked to indicate the time the sample deployment started and ended, and the time they spent outside and with their windows open. Participants were instructed to wear their silicone wristband sample 24 hours a day during the designated sampling week, including while sleeping, eating, and showering, consistent with other studies (Samon et al., 2022). In total, 49 participants completed questionnaires. The design of this study was intended to capture the PAH concentrations our participants would normally encounter in their daily lives; this includes indoor, outdoor, home, and workplace environments, to explore the potential influence of behaviors and housing characteristics on total personal exposure to vapor-phase PAHs.

2.4. Participant Compliance

Of the 49 enrolled participants that completed questionnaires, 46 participants (31 households) returned at least one sample. Of those 46 participants, 7% completed two sampling periods, 30% completed three sampling periods, and 63% completed all four sampling periods. Table S3 details samples mailed out and received in each sampling period and Table S4 summarizes participant compliance and conditions of sample receipt. 96% of samples analyzed were deployed in an acceptable time frame and packaged correctly

2.5. Materials

Chemical standards were purchased from Accustandard (New Haven, CT, USA), Chiron (Trondheim, Norway), and Sigma-Aldrich (St. Louis, MO, USA). Deuterated surrogates were acquired from CDN Isotopes (Pointe-Claire, Quebec Canada). Solvents used were Optima-grade or equivalent (Fisher Scientific, Pittsburgh, PA, USA).

2.6. Silicone Wristband Passive Samplers

2.6.1. Preparation for deployment and extraction

Silicone wristbands were purchased from 24hourwristbands.com (Houston, TX, USA) and prepared as previously described (Anderson et al., 2017; Donald et al., 2016a; O’Connell et al., 2014). From each conditioned lot, wristbands were tested for elasticity and analyzed using GC-MS to ensure that there was sufficient reduction of oligomers for analytical sensitivity (Dixon et al., 2019). Conditioned wristbands were stored at 4°C in airtight metal containers until deployment.

Following deployment, silicone wristbands were rinsed with isopropyl alcohol to remove surface particles, extracted with two rounds of dialysis with ethyl acetate, and underwent solid phase extraction as previously described (Anderson et al., 2017; Dixon et al., 2019; O’Connell et al., 2014). Solvent was exchanged to iso-octane, and sample aliquots were stored at 4°C prior to instrument analysis.

2.6.2. Quality assurance and quality control

Quality assurance methods and quality control samples are detailed in Table S5. Background correction by batch was performed with results from blank quality control samples by batch. Laboratory processing blanks consistently had the greatest number of PAH detections and concentrations of any other blank sample type, including a mailed trip blank, and were relied on for each month’s background correction, as detailed in Table S6.

2.7. PAH Analysis

Silicone extracts were analyzed for 62 parent and alkylated PAHs, quantified with an Agilent (Santa Clara, CA) gas chromatograph (7890 GC; select PAH column, 30 m×250μm×0.15μm) and triple quadrupole mass spectrometer (7000 MS/MS) as previously reported (Anderson et al., 2017; Anderson et al., 2015; Dixon et al., 2018; Donald et al., 2019; Minick and Anderson, 2017; Paulik et al., 2018). GC-MS/MS data was analyzed using MassHunter Quantitative Analysis v.B.06.00 SP1 build 6.0.388.1 software (Agilent Corp. Wilmington, DE, USA). Target analyte concentrations were surrogate corrected; extraction surrogate recoveries are summarized in Table S7. Instrument parameters and physiochemical properties of analytes are reported in Tables S8 and S9. Perylene-d12 was used as the internal standard.

2.8. Data analysis

All data processing, analysis and visualization were performed in R (version 4.4.3).

2.8.1. Chemical data preparation

Four PAH analytes were excluded from all data analyses due to poor calibration performance (benz[a]anthracene and co-eluents benz[j] and [e]aceanthrylene) or persistent matrix interference (2,6-dimethylnaphthalene and 2,6-diethylnaphthalene). PAH concentrations from silicone wristbands were reported as nanogram per gram of silicone (ng/g), where the silicone mass per wristband is the average mass of three wristbands from the conditioning lot (reported in Table S10). All concentrations were Log10-transformed.

2.8.2. Questionnaire data preparation

Multiple choice questions with no established order were treated as categorical variables, and ordinal responses that indicated increasing quantity or frequency were analyzed as numeric variables. Missing values were addressed on a case-by-case basis, and some variables were withheld from analysis due to insufficient responses. In some instances, missing values were supplemented with answers from participants in the same household, or the median or majority response was imputed. Multicollinearity of variables was assessed using Cramer’s V values, and in cases of high correlation between variables, only one was retained for analysis (Figure S4) (Cramer, 1946). More complete information on questionnaire data handling is available in Table S11. Data for 25 total questionnaire variables are summarized in Table S12.

2.8.3. Statistical Modeling

Shared information between PAHs was calculated via Spearman correlation. Questionnaire data and chemical exposure data from wristbands were integrated in a series of quantitative (i.e., concentration-based) and binary (i.e., detection-based) generalized linear regression models to interpret questionnaire variable importance in predicting PAH exposure. Quantitative exposure models for individual chemicals were evaluated with the Log10-transformed concentration of values above the LOD as the outcome variable.

Quantitative models were limited to chemicals with at least 25% of observed values above the limit of detection (LOD). Individual chemicals had a wide range of detection frequencies, and quantitative chemical model results, and the detection rate directly impacts the number of quantitative measures included in the model. Quantitative model results for chemicals with lower detection rates may be interpreted with this context, and binary or PC model results may be more useful in those cases. To ensure model convergence, binary models were limited to chemicals with between 20 and 140 detections (n=162). Frequency of detections for specific chemicals is detailed by month in Table S13; a total of 25 and 21 PAH analytes met these criteria for the quantitative and binary models, respectively.

Projection pursuit principal component analysis (PCA) was run on filtered quantitative concentration data (25 analytes), which was centered and scaled (Stacklies et al., 2007). The benefit of the projection pursuit approach is that it does not require imputation of values below the LOD. The outcome of the PCA was used for two high-level exposure models, considering the first two principal component scores for each wristband as outcome variables.

All models were fit using the glmnet version 4.1-8 R package (Friedman et al., 2010; Stanfill et al., 2019). The 25 questionnaire variables plus sampling month (26 total variables listed in Table S14) were used as fixed effect explanatory variables, and a random effect associated with individual participants was specified to account for correlation among multiple measurements from the same individual., No interactions among variables were included in the generalized linear models. Quantitative concentration models were fit using a mixed effects linear model with a lasso penalty for variable selection (Tay et al., 2023) and were fit to each high-level (PC1 and PC2) and individual chemical model. Binary detection models were fit using a mixed effects linear model with a conditional binomial distribution and lasso penalty (determined with leave-one-out cross (LOO) validation). Model performance was quantified using the R² value between observed and predicted Log10 concentrations and Area Under the Curve (AUC) for the detection models. Models with a fitted R² greater than 0.60 or an AUC greater than 0.70 for detection models (and greater than the LOO AUC) were evaluated further. Twelve individual analytes were evaluated further in concentration models, and ten in the detection models (Figure S5).

Importance of retained variables was measured by the decrease in R² when the variable was omitted from the model. Variable importance values were scaled from zero to one. The characterization of the mean effect of variables on the response variable of interest, based on the fitted model and given other variables retained in the model, were evaluated by calculating accumulated local effects (ALE) values across the range of variable values. The change in predicted response variable as the explanatory variable increases (or a “yes” answer compared to a “no” answer for categorical data) was calculated and is referred to the “ALE slope” hereafter (Apley and Zhu, 2020).

3. Results and Discussion

In 160 wristband samples collected across eight months, a total of 50 PAHs were detected in at least one sample, 17 of which were alkylated PAHs. Phenanthrene, dibenzothiophene, and 2-methylphenanthrene were detected in all samples (Table S13; Figure 1). Pyrene, 2-methylnaphthalene, fluoranthene, retene, fluorene, and 1-methylphenanthrene were detected in all but one or two samples. Eight PAHs were not detected in any sample. Two participants in the first sampling group had notably elevated PAH exposures in terms of detection frequencies and concentrations throughout all sampling timepoints. When these samples were not considered, 41 PAHs were detected at least once in the remaining samples. Full distributions and summary data of individual PAH concentrations by month are reported in Figure S6 and Table S13 respectively. Of the chemicals detected in at least 50% of samples in each month, phenanthrene accounted for the largest median proportion (>15%) across all months, and the relative proportions for many PAHs were variable (Figure S7).

Figure 1.

Detection frequencies (A) and median concentrations in ng/g (B) of PAHs detected in at least one sample. Columns represent months of sampling in chronological order.

The distributions of responses to each questionnaire variable as used for data analysis are detailed by sampling group and for all participants in Table S12 and a full summary of participant demographic data is provided in Table S15. Notable observations that were not captured in further data analysis are time spent outside and time with windows open for each sampling period (Table S16 and Figure S8) due to poor response rates for these questions (77%; Table S4). Responses received indicated that the first sampling group had a greater proportion of individuals spending time outside and opening windows, and across both groups, time outside and with windows open increased with temperatures, as in May and June. While not included in subsequent models, these differences in exposure to outdoor air and sources in different months likely influenced measured individual PAH exposures and were qualitatively considered in the interpretation of modeling results below. Also not included in the following models were variables related to outdoor temperature (Table S1, Figure S3), precipitation and PM_2.5 (Table S2), and wind velocity (Figure S2), as this information was assumed to be captured by the “month” variable included in models. Bramer et al., discussed study-specific environmental conditions in St. Helens more extensively (Bramer et al., 2025).

3.1. Correlation of Individual PAHs

Individual PAHs co-occurred, as demonstrated with significant correlations between exposure to individual PAHs. PAHs with similar chemical structures were highly correlated. Hierarchical clustering identified strong correlations within (1) retene, a three-ring alkylated PAH, (2) all two-ring PAHs, (3) three to five ring PAHs, and (4) another three-ring cluster with primarily alkylated phenanthrenes (see boxes in Figure 2). As an example of correlation strength among structural groups, all alkylated naphthalenes are strongly correlated with each other and cluster together (r > 0.73). Only a few chemicals had insignificant correlations with other PAHs, predominately retene and benzo(c)fluorene (p < 0.05).

Figure 2.

Pairwise PAH correlation of log10 transformed concentrations; each comparison has at least five complete paired observations. Large dark blue circles indicate strong positive correlation. Blank squares were insignificant based on Spearman correlations (p > 0.05). PAHs were organized with hierarchical clustering; the four strongest clusters are bounded with black rectangles.

3.2. Mixed Model Results for Aggregate PAH Metrics

Projection pursuit PCA for twenty-four PAHs yielded one principal component (PC1) which accounted for 57.2% of variability in the data, and a second principal component (PC2) which accounted for 22.0% of variability in the data (79.2% total variability accounted for). Loadings for each PAH were entirely positive for PC1 (Table S17) and increasing PC1 scores were interpretable as a surrogate measure for increasing total PAH exposure. The loadings for PC2 separated out into two groups of chemicals: 2-ring PAHs (i.e. parent and alkylated naphthalenes) all had positive loadings, while 3-ring or more PAHs all had negative loadings. PC2 was interpreted as an indication of the ratio between 2-ring and 3 or more ring PAHs. Use of PC1 and PC2 as metrics to represent (1) total PAH exposure increases and (2) differential exposure to 2-ring PAHs versus 3 or more ring PAHs, respectively, allowed for a first pass data analysis involving fewer chemical variables. This simplification of the data was further corroborated by Figure 2, where individual PAHs were positively correlated, and ring-sizes were distinguished with hierarchical clustering.

Two generalized linear mixed effects models were constructed with PC1 and PC2 scores as response variables. These provided a good model fit; p-values for a likelihood ratio test were 3.86x10⁻⁹ and 5.49x10⁻¹¹ respectively and the R² values for the PC1 and PC2 scores as outcome variables were 0.5596 and 0.5385, respectively. A full list of variable importance scores is included in Table S14.

The generalized linear mixed model with PC1 scores as the response variable identified flooring type, participant age, and sampling month as the strongest predictors of total increased PAH exposure (Table 2). The PC2 score model indicated that differential exposure to naphthalenes, relative to 3 or more ring PAHs, was linked with floor type, operation of heavy machinery, use of a wood stove, and whether an odor was noted by the residents around their house (Table 2). A characteristic to note about the dataset with the ‘Odor’ variable is that two participants in sampling Group 1 with some of the highest overall PAH exposures represented two out of a total of six participants that reported residential odors.

Table 2.

Principal component model’s variable importance values for select predictors. Variables listed in order of relative importance, with the highest importance scores (greater than 0.2) for each PC model shaded in orange. Importance scores were scaled from zero to one. Only variables with scores greater than 0.1 are included in this table.

PC1 Model (total PAH exposure)		PC2 Model (differential exposure to 2-ring PAHs)
Variable	Importance (Scaled)	Variable	Importance (Scaled)
Hardwood	1	Laminate	1
Carpet	0.765	Carpet	0.533
Month	0.447	Machinery	0.367
Laminate	0.416	Wood Stove	0.300
Age	0.204	Hardwood	0.291
Odor	0.135	Odor	0.210
Grill	0.133	House Age	0.199
AC	0.116	Month	0.124

Based on both PC models, flooring variables are important predictors of total exposure, and differential exposure to two ring PAHs versus three or more ring PAHs. Flooring variables, along with other notable variables are further investigated with individual chemical models and empirical comparisons in Section 3.4.

3.3. Mixed Model Results for Individual PAHs

To follow up on the more general PC model analysis, individual PAHs were investigated in relation to questionnaire variables using generalized linear mixed models. Results from these individual PAH models are presented in Figure 3A for binary detection-based data and Figure 3B for quantitative concentration data. These figures highlight which questionnaire variables were identified as important predictors of exposure to the individual PAHs based on retention in the model and variable importance. As discussed by Bramer et al. (2025), the utilization of both concentration and detection models provides a full picture of PAH activity across timepoints and between participants (Bramer et al., 2025).

Figure 3.

Individual chemical (A) binary detection model and (B) quantitative concentration model: retained variables and importance. Scaled values (0-1; blue to red) correspond to R² reductions when the specified questionnaire variable is omitted from the linear mixed effects model. The nearer to one a value is, the more impact a variable has on the detection of the indicated PAH. Grey boxes indicate that the variable was not retained in the model for the specified chemical.

For the detection model, the month variable was retained in all chemical models, and in all but two chemical models (benzo(j)fluoranthene and benzo(k)fluoranthene), had the highest scaled importance score among variables (Figure 3A). There were no other retained variables with scaled importance scores greater than 0.5 for 2-ring PAHs or acenaphthylene, although ‘odor’ for 1,2-dimethylnaphthalene was close. Acenaphthene had relatively high importance scores for predictor variables of whether participants had a wood stove and education level. Benzo(b)fluoranthene, benzo(j)fluoranthene, and benzo(k)fluoranthene had relatively high importance scores for the predictor variables of race. Interestingly, the benzo(b)fluoranthene model identified several combustion and smoke sources including bonfires, grilling, and backyard burning as important predictors.

For the concentration-based model, month was similarly retained in all chemical models and had high scaled importance scores among variables (Figure 3B). All three types of flooring in houses that were investigated, laminate, hardwood, and carpet, were retained and identified as relatively important variables in predicting quantitative individual PAH exposure for many PAHs. Anthracene and benzo(e)pyrene models also indicated that odor reported around a participants’ residence was a strong predictor of concentration.

3.4. Important Predictor Variables

3.4.1. Sampling Month

There are multiple lines of evidence with the PC1 model for total PAH exposure and numerous individual chemical models derived from both detections and concentrations, all of which agreed month of sampling was an important predictor variable of PAH exposure. There were some observable changes in median sum PAH concentrations grouped by structural characteristics over time, specifically for 2-ring PAHs, with the highest median sums in spring and winter months, and lowest in fall and summer months (Figure S9). Additional investigation into empirical changes in individual PAH exposure across time was conducted to better understand the assigned importance of this variable in the mixed models.

In terms of PAH detections, presence of some chemicals varied considerably across months in (Figure 1, Table S13). Acenaphthene, for example, was detected in approximately 50% of samples in November and March and in less than 25% of samples in December through February, and June. In general, there were fewer PAHs detected in 50% of samples or more in May (15 PAHs) than any other month, compared to February or April (21 PAHs), which had the most.

To further investigate changes in individual PAHs across time, PAHs retained in the individual chemical models were plotted in a heatmap within ‘seasons’ (Figure 4) which represent two sequential months, including data from both cohorts to avoid the potential confounding factor of participant group. Grouping months of data into seasons accounted for potential differences between sampling groups. Data from consecutive months were compared as ‘seasons’ with November and December data assigned to fall, January and February were assigned to winter, March and April were assigned to spring, and May and June were assigned to summer. These seasonal bins represent not only potential changes in outdoor conditions but potential changes in behavior. For example, across seasons, there are differences in time spent outside captured with monthly questions in this study and expected changes in activities dependent on planting and yardwork, school schedules or holidays, which could influence time at home and behaviors.

Figure 4.

Seasonal comparisons of Log10 concentration (ng/g) data, scaled by row to examine temporal differences of aggregate cohort data within a single chemical.

As suggested by the PC models and individual chemical models previously discussed, there are temporal trends in PAH exposure data across the repeated measures in this study. Additionally, these trends are chemical-dependent, and as discussed with the correlation plot in Figure 2 and principal component analysis, similar chemicals, especially by molecular weight, tend to cluster (Figure 4). In general, winter and summer months were associated with the highest concentrations. Whether winter or summer months had higher concentrations of PAHs appeared to change with chemical features, including molecular weight, with some exceptions. Interestingly, a group of alkylated naphthalenes (1,2-dimethylnaphthalene, 2-methylnaphthalene, 1,5-dimethylnaphthalene, and 1,6 and 1,3-dimethylnaphthalene) were relatively low in summer, while concentrations of several three and four ring PAHs were highest in summer (3,6-dimethylphenanthrene, 1-methylpyrene, pyrene, and retene). Notably, spring is the season with the lowest PAH exposure for almost all of the highest molecular weight compounds.

Potential factors that could help explain these temporal trends are seasonal changes in both indoor and outdoor air quality, or seasonal changes in behavior that contribute to or reduce PAH exposure. There were relatively high precipitation totals in the combined fall and spring months, which may have reduced outdoor PAH concentrations, including high molecular weight PAHs specifically. There are also indoor and outdoor sources of PAHs that tend to be more prevalent in winter or summer months, relative to fall and spring. Retene is a good example of a PAH with seasonal variability in outdoor air, with unique sources. This PAH is found in resinous woods; as a result, smoke generated from wildfires or other wood-combustion processes can contain high levels of retene (Miller et al., 2017; Navarro et al., 2019). In this dataset, retene did not correlate well with other PAHs found in wristbands in this study, and there is a temporal difference in exposure with elevated concentrations in May and June 2023. During these months, there were several large wildfires burning to the North, so it is possible that long range transport of the wildfire smoke was impacting the air quality for St. Helens residents in these months. There was not a relative increase in AQI during May and June compared to other months, so this may reflect other more local unidentified wood-fuel combustion sources (e.g. recreational fire pits, brush burning, etc.). In addition to wood burning and retene, other PAH sources in May and June may include diesel-powered tools used for yardwork, such as lawn mowers, or grilling. In terms of seasonal behaviors, it is important to consider the fact that people spend more time outside and with their windows open in the warmer months, so indoor sources of PAHs may have less of an impact when compared to winter months. Residents indicated they used a fireplace or wood burning stove more frequently with lower temperatures in the winter, which could contribute to elevated PAH exposure in the winter, for example with the naphthalenes.

3.4.2. Flooring

For the binary models, carpet was retained in seven of ten chemical models and hardwood in six of ten; laminate flooring was not retained. A heatmap illustrating relative concentrations (scaled by chemical) among floor types is shown in Figure 5. All 2-ring PAHs were in the highest relative concentrations when hardwood floor was indicated in the participant’s house. Interestingly, this was also true of retene, even though all other 3 or more ring PAHs were in the highest relative concentrations with laminate flooring or carpet (just for acenaphthylene).

Figure 5.

Flooring type comparisons of mean Log10 concentrations (ng/g) of retained variables in the individual chemical models, scaled by row.

There is some literature to suggest that flooring type influences indoor air concentrations of pollutants. Carpet, for example, has been proposed to act as a high surface area sink for semi-volatile compounds, which can lend to sorption and re-emission in the house over time. Physiochemical properties, such as molecular weight, could also influence carpet sorption or re-emission (Becher et al., 2018; Noorian Najafabadi et al., 2022). Flooring type is also related to cleaning in the home, which could in turn influence ambient indoor PAH levels. Flooring materials themselves may also be a chemical source, for example plasticizers or polishes in flooring could contribute SVOCs to indoor environments (Guan et al., 2024). A recent review from Landeg-Cox, Middleton et al., from 2025 on building material emissions of SVOCs also noted in their conclusions that there is also not much research on SVOC emission rates from building materials, including flooring materials (Landeg-Cox et al., 2025; Zhang et al., 2025). We identified very limited research related to flooring impacts on PAH concentrations specifically in indoor environments. One paper identified that addressed flooring as a direct source of PAHs was a biomonitoring study related to parquet glue, which contains coal tar, used for hardwood flooring in residences built in the 1950s in Germany (Heudorf and Angerer, 2001).

However, this study found no detectable differences with biomonitoring between residents living in parquet glue flooring containing residences, versus those that did not. In a chamber study, 2-methylnaphthalene, and naphthalene were two of the main SVOCs determined to be emitted from PVC floorings (Guan et al., 2024). Given the importance of this variable across models coupled with limited available research related to PAHs and flooring as a source in indoor environments, the influence of floor types on indoor chemical exposure warrants future investigation and inclusion in all exposure questionnaires, supplemented with questions on cleaning methods and frequency.

3.4.3. Other Variables

Use of wood stoves and participant age were also retained in multiple single chemical models, both binary and quantitative, and were highlighted in both PC models. Of note, wood stoves were only retained in models for 3-5 ring PAHs, which aligns with relative importance of wood stove in the PC2 model for differentiating between PAH structures, and not the PC1 model indicative of total PAHs. Age was retained in models for a variety of PAH ring sizes, which also aligns with the variable importance in the PC1 model, but not the PC2 model.

The predicted ALE slope for age of the participant (Age) in the PC1 model was positive, which indicated that the older a participant is, the more likely they were to have higher total PAH exposure. Participants in the age range of 36-65 years old may have occupations or daytime activities that might be more likely to expose them to PAHs, while people under 18 are most likely in school. Most of the participants in the 66 and older range indicated that they were retired and spent most of their time at home. Age likely corresponds with different environments and behaviors that inform exposure.

While not highlighted in the PC models, smoking variables in relation to PC1 and PC2 indicated that smoke exposure from cigarettes or cigars, especially indoors, resulted in increased PAH exposure and specifically increased PAH exposure to naphthalenes. Smoking scores were also retained in most individual binary and quantitative chemical models, as supported with existing literature (Cocco et al., 2007; Lu and Zhu, 2007; McLarnan et al., 2024).

Another consideration not explored with the current models is geographic factors such as home or work location in relation to PAH sources that may impact exposure for certain participants. In some cases, proximal point sources were identified closest to addresses of relatively high-exposure individuals. Future work will consider the influence of spatial relationships to chemical exposure in addition to behavioral and demographic factors. More broadly, exposure studies should incorporate questionnaires and a dataset of proximal point sources to provide more complete context for measured exposure concentrations.

3.5. Repeated measures

This study allowed for characterization of the variability in the repeated measures of an individual’s PAH exposure. Distributions of variability for repeated measures from individuals for frequently detected PAHs are shown in Figure S10 and discussed in more detail in Bramer et al., 2025. Briefly, variability between participants was greater than within an individual’s repeated measures, and some participants had much more consistent exposures over time than others across repeated measures. Across samples, alkylated naphthalenes were the most variable with coefficients of variation (COV) over 500% while PAHs with four or more rings tended to be less variable (COV <150%) (Figure S10).

The design of this study is similar to a couple previously published studies in terms of collecting questionnaire data with silicone wristbands over repeated sampling periods (Donald et al., 2016b; Guo et al., 2021). Guo et al (2021) in particular collected behavioral information with PAH concentrations and found a similar trend in naphthalene presence in winter months and a link with indoor environment characteristics to PAH exposure. Unlike other current silicone wristband repeated measure studies, this addresses multiple deficiencies identified by Murcia-Morales et al., including reproducibility and variability quantification, sufficient sample size (>35 participants), and administration of surveys related to participant behaviors (Murcia-Morales et al., 2024).

3.6. Limitations

Care should be taken when interpreting differences in concentrations between chemicals due to differences in uptake rates into the silicone polymer. Low molecular weight PAHs, such as naphthalenes, have faster uptake rates and time-to-equilibrium than heavier compounds. The majority of detections made in the 7-day deployment described here are expected to still be within the window for linear uptake (O’Connell et al., 2022); (Samon et al., 2024). Additionally, comparing changes in relative contributions of PAHs between participants or other explanatory variables is appropriate and relevant. The use of silicone wristbands as personal passive samplers approximates the bioavailable, vapor-phase, fraction of a chemical that could be inhaled or absorbed dermally but does not account for ingestion exposure (Cocco et al., 2007).

Baseline questionnaire data was collected ahead of sampling and reported behaviors may have differed from actual behaviors during sampling, further, responses to monthly questionnaires had limited response rates. The authors examined potential correlations between questionnaire variables, but there may still be relationships between variables that exist and impact models and interpretations in this paper (e.g. collinearity or interactions).

4. Conclusion

This study contributes valuable repeated PAH exposure measurements to a limited body of literature integrating personal PAH exposure with individual behaviors and environments. The results demonstrate through correlation of exposure factors and repeated PAH exposure measures that questionnaire data can inform exposure data, and combining the two datasets can provide more nuanced insight into behaviors and variables that influence exposure. The number of repeated measures in this personal silicone passive sampling study is also unique and lends insight into personal exposure variability. The results demonstrated that exposure to naphthalenes was distinct from higher molecular weight PAHs in terms of temporal trends and relationship to exposure factors. Multiple models incorporating principal component metrics, detection and concentration data were leveraged to elucidate driving factors of total PAH exposure and differential exposure between 2-ring and larger PAHs. Floor type, time period, and combustion-related activities including use of wood-burning stoves and heavy machinery use were implicated in personal PAH exposure increases and increased prevalence of naphthalenes through multiple lines of evidence. Based on the results of this study, exposure scientists should distribute surveys and include questions related to important exposure factors identified in this study (e.g. flooring types in indoor environments). Identification of important personal exposure factors across time and PAH types in this study informs and invites future study of specific personal exposure sources and reduction strategies.

• Personal exposure assessments benefit from collection of questionnaire data

• Silicone wristbands enabled collection of repeated measures with high compliance

• Repeated measures yield variability metrics and better represent personal exposure

• Important determinants of PAH exposure included month and flooring types