Urinary biomarkers of polycyclic aromatic hydrocarbons in pre- and peripubertal girls in Northern California: Predictors of exposure and temporal variability

Abstract

Background

Polycyclic aromatic hydrocarbons (PAHs), a class of chemicals produced as combustion by-products, have been associated with endocrine disruption. To understand exposure in children, who have been less studied than adults, we examined PAH metabolite concentrations by demographic characteristics, potential sources of exposure, and variability over time, in a cohort study of pre- and peri-pubertal girls in Northern California.

Methods

Urinary concentrations of ten PAH metabolites and cotinine were quantified in 431 girls age 6–8 years at baseline. Characteristics obtained from parental interview, physical exam, and linked traffic data were examined as predictors of PAH metabolite concentrations using multivariable linear regression. A subset of girls (n = 100) had repeat measures of PAH metabolites in the second and fourth years of the study. We calculated the intraclass correlation coefficient (ICC), Spearman correlation coefficients, and how well the quartile ranking by a single measurement represented the four-year average PAH biomarker concentration.

Results

Eight PAH metabolites were detected in ≥ 95% of the girls. The most consistent predictors of PAH biomarker concentrations were cotinine concentration, grilled food consumption, and region of residence, with some variation by demographics and season. After adjustment, select PAH metabolite concentrations were higher for Hispanic and Asian girls, and lower among black girls; 2-naphthol concentrations were higher in girls from lower income households. Other than 1-naphthol, there was modest reproducibility over time (ICCs between 0.18 and 0.49) and the concentration from a single spot sample was able to reliably rank exposure into quartiles consistent with the multi-year average.

Conclusions

These results confirm diet and environmental tobacco smoke exposure as the main sources of PAHs. Controlling for these sources, differences in concentrations still existed by race for specific PAH metabolites and by income for 2-naphthol. The modest temporal variability implies adequate exposure assignment using concentrations from a single sample to define a multi-year exposure timeframe for epidemiologic exposure-response studies.

1. Introduction

Polycyclic aromatic hydrocarbons (PAHs) are a class of ubiquitous chemicals generally produced as combustion by-products; of these chemicals, only naphthalene is produced commercially in the United States (ATSDR, 1995, 2005). People are usually exposed to a complex mixture of PAHs rather than individual chemicals (CDC, 2009). Non-occupational exposure sources include vehicle exhaust, residential heating sources, smoke from wood, coal, or gas, cigarette smoke, and grilled foods (ATSDR, 1995). PAHs are absorbed by the body (i.e., skin, respiratory tract, and gastrointestinal tract), and can be metabolized into monohydroxy-PAHs (OH-PAHs), and eliminated in the urine or feces within a few days (Chetiyanukornkul et al., 2006; Ramesh et al., 2004). Urinary OH-PAHs have therefore been used as biomarkers for assessing recent exposure to PAHs because biomonitoring accounts for all possible routes (e.g., inhalation, diet) (Aquilina et al., 2010; Li et al., 2010a; Nethery et al., 2012; Scherer et al., 2000). Metabolism of PAHs differs among exposure sources and varies by person (Jacob and Seidel, 2002; Li et al., 2012b, 2016; Lin et al., 2016).

Previous studies have detected OH-PAHs in children not exposed to residential biomass fuels, not near occupational sites, nor smoking (CDC, 2017; Health Canada, 2015; Hemat et al., 2012; Thai et al., 2016; Wilhelm et al., 2008). However, only a handful of studies have quantified an array of PAH urinary metabolites in children to assess sources of exposure (Alghamdi et al., 2015; Jung et al., 2014; Kang et al., 2002; Yoon et al., 2012). Many studies solely measured 1-hydroxypyrene (Cavanagh et al., 2007; Fiala et al., 2001; Freire et al., 2009; Martinez-Salinas et al., 2010; Morgan et al., 2015; Mucha et al., 2006; Ochoa-Martinez et al., 2016; van Wijnen et al., 1996), which is not representative of the mixture of PAHs in general environmental exposures.

Limited published data exist on the temporal variability of OH-PAH urinary concentrations. A recent study found that a person's PAH metabolite concentrations in urine spot samples, first morning voids, and 24-h voids have a high degree of correlation (Li et al., 2010b). Time intervals of a few days (Fiala et al., 2001) or seasons (Peters et al., 2017; van Wijnen et al., 1996) have previously been studied. Data on the temporal variability of OH-PAHs across years, especially among children, are sparse (Jung et al., 2014) but important because health outcomes are associated with exposures over time intervals longer than the few days of metabolite elimination. We used a Breast Cancer and the Environment Research Program (BCERP) cohort to evaluate demographic differences, potential sources, and temporal variability of OHPAH urinary concentrations among pre- and peri-pubertal girls in Northern California.

2. Methods

2.1. Study population

The BCERP cohort is comprised of over 1200 girls who were enrolled in 2005–2006 into a longitudinal study of puberty at three U.S. sites: Mount Sinai School of Medicine, Cincinnati Children's Hospital, and Kaiser Permanente Northern California (KPNC) (Biro et al., 2010; Hiatt et al., 2009). Eligibility criteria included female sex, 6–8 years of age, and no underlying endocrine-associated medical conditions. This analysis included only girls recruited from the KPNC Health Plan in the San Francisco Bay Area, as other sites did not measure PAHs. The site obtained informed consent from a parent or guardian and child assent, approved by KPNC's institutional review board (IRB) with the Centers for Disease Control and Prevention (CDC) reliance upon the KPNC IRB.

2.2. Data collection

Demographics, anthropometry, and urine were obtained at baseline and annually thereafter. The girl's parent or guardian completed an annual questionnaire, including girl's race/ethnicity, diet, other demographic variables, and residential history. The baseline variables were used for this analysis. The girl's race/ethnicity was classified in the following priority order: black (regardless of ethnicity), Hispanic (any race other than black), Asian or Pacific Islander (non-Hispanic), and white (non-Hispanic) (hereafter referred to as black, Hispanic, Asian, and white). The parent or guardian was asked for the total, pre-tax income from all income sources by all family members in the child's household in the past year, using a series of questions to determine the income group: less than $12,000, $12,000- < $25,000, $25,000- < $50,000, $50,000- < $75,000, $75,000- < $100,000, $100,000 or more. For this analysis, household income was categorized as < $50,000, $50,000- < $100,000, and ≥$100,000. Recent grilled food consumption was defined by a “Yes” response to the question “During the past two days, has (child's name) eaten any foods that were cooked on a grill or barbeque? Please do not include foods that may have been cooked on an electric grill such as a George Foreman grill.” Traffic metrics have been described previously (McGuinn et al., 2016). Briefly, baseline residential addresses (2005–2006) were geocoded using the California Environmental Health Tracking Program's (CEHTP) geocoding tool, and then the address was linked to traffic exposure data (California Department of Transportation Highway Performance Monitoring System data from 2004) using the CEHPT's Traffic Volume Linkage Tool. Distance-weighted traffic density (highest average annual daily traffic (AADT) with Gaussian weight based on distance from residence) was calculated within a 150m buffer zone around each geocoded address. Using residential city, the girl's region of residence was classified as the San Francisco Peninsula (i.e., the City and County of San Francisco, San Mateo County), North Bay (i.e., Marin County, Sonoma County), or East Bay (i.e., Alameda County, Contra Costa County, San Joaquin County, Stanislaus County, Solano County).

Height and weight were assessed at annual clinic visits by study staff who had been uniformly trained and certified to use calibrated scales and stadiometers. Body mass index (BMI) was calculated as weight/height-squared and then classified by age- and sex-specific BMI percentiles based on the 2000 growth charts from the CDC (CDC, 2002). Categories of baseline BMI were defined as normal (below the CDC age and sex-specific 85th percentile) and overweight (above the 85th percentile). Spot urine samples were collected annually. The season of sample collection, based on the date of the baseline urine sample, was classified as: spring (March–May), summer (June-August), fall (September– November), or winter (December–February).

2.3. Quantification of biomarkers

Girls with sufficient baseline urine sample volume for PAH biomarker measurements were included in this study (N = 431). In addition, samples from a subset of 100 girls with stored urine available at both the second and fourth years of the study (year 2 and year 4, respectively henceforth) were selected to assess temporal variability of PAH biomarkers concentrations. Urinary concentrations of ten OH-PAH metabolites (i.e., 1-naphthol, 2-naphthol, 2-hydroxyfluorene, 3-hydroxyfluorene, 9-hydroxyfluorene, 1-hydroxyphenanthrene, 2-hydroxyphenanthrene, 3-hydroxyphenanthrene, 4-hydroxyphenanthrene, 1-hydroxypyrene) as well as urinary cotinine, a nicotine metabolite, were measured at the National Center for Environmental Health laboratories at the CDC using previously published methods (Bernert et al., 2005; Li et al., 2006, 2014). Briefly, the PAH metabolite conjugates in urine were hydrolyzed enzymatically and extracted using pentene through liquid-liquid extraction; the extracts were evaporated and the PAH metabolites derivatized. The baseline samples were quantified using gas chromatography isotope dilution high-resolution mass spectrometry (Li et al., 2006); annual samples thereafter were quantified using gas chromatography isotope dilution tandem mass spectrometry (Li et al., 2014). Quality control and quality assurance checks have been previously described (Li et al., 2006, 2014). Urinary cotinine was quantified by high performance liquid chromatography/atmospheric pressure ionization tandem mass spectrometry; each analytical run included one water blank and two quality control samples (Bernert et al., 2005; McGuffey et al., 2014). Two girls were missing 1-hydroxypyrene measurements; 12 girls were missing cotinine measurements. Creatinine was also measured at CDC using the Roche Creatinine Plus Assay (Roche Diagnostics, Indianapolis, IN) on a Roche Hitachi 912 Chemistry Analyzer (Hitachi Inc., Pleasanton, CA).

2.4. Statistical analysis

Statistical analysis was conducted using SAS (version 9.3; SAS Institute Inc., Cary, NC), with a focus on biomarkers detected in > 60% of samples. Biomarker concentrations were non-normally distributed so geometric means (GMs) were calculated, with measurements below the limit of detection (LOD) imputed using LOD/√2 (Hornung and Reed, 1990). Study log-transformed concentrations were compared to those from the U.S. National Health and Nutrition Examination Survey (NHANES) using two-sample t-tests (Armitage and Berry, 1987). NHANES is a cross-sectional, nationally representative survey of the U.S. noninstitutionalized civilian population (CDC, 2017). In the NHANES 2005–2006 cycle, PAH metabolites were measured in spot urine samples from one-third of subjects over the age of 6 years old for a total (both sexes) of about 340 6–11 year olds (CDC, 2017).

We assessed correlations among covariates and analytes using Spearman's method. Values for three fluorene (2-hydroxyfluorene, 3-hydroxyfluorene, and 9-hydroxyfluorene) and three phenanthrene (1-hydroxyphenanthrene, 2-hydroxyphenanthrene, and 3-hydroxyphenanthrene) metabolites were summed (labeled Σfluorene and Σphenanthrene, respectively). To analyze demographic differences in OH-PAH concentrations, we used PROC GENMOD to run a demographics-only model with age, race/ethnicity, income, BMI, and ln (creatinine). Because few girls were age 8 years at baseline, age was dichotomized as 6–6.9 years old or 7 years old or above. Household income and parent/guardian education level were significantly correlated with each other so only income was used in multivariable models. BMI did not improve model fit so it was removed from the final models. To examine whether sources of PAH exposure could explain the demographic differences, we used PROC GENMOD to conduct multivariable regression analysis adjusting for age, race/ethnicity, income, and ln(creatinine). Proportional change in urinary metabolite concentrations, expressed as the exponentiated beta coefficient(e^β) from the multivariable regression, was calculated to compare biomarker concentrations among girls with an exposure or covariate of interest to girls in the reference category. Cotinine concentrations were non-normally distributed. To ensure that an extreme value was not highly influential, concentrations were log-transformed and the multivariable models were re-run with outliers removed (any observations more than three interquartile ranges below the 25th percentile or above the 75th percentile) to evaluate the influence on the regression analysis.

To assess temporal variability of the OH-PAHs and the sums, Σfluorene and Σphenanthrene, we calculated Spearman correlation coefficients, intraclass correlation coefficients (ICC) from a one-way random-effects ANOVA model (Hertzmark and Spiegelman, 2010), and performed a surrogate category analysis (Hauser et al., 2004; Teitelbaum et al., 2008) to assess how well quartile ranking by a single biomarker measurement (“surrogate sample”) represented the four-year average OH-PAH concentration (over the three visits, i.e. baseline, year 2, and year 4). All OH-PAH concentrations were natural-log transformed and then back transformed to obtain geometric means. A multiyear mean concentration (mean of the three repeated measures over the four-year period) was calculated for each girl to represent her four-year average OH-PAH exposure. Quartile cut points were created using the distribution of each “surrogate sample” (i.e., baseline, year 2, and year 4). We then averaged the four-year mean concentrations of girls within quartiles of each “surrogate sample.” A monotonic increase in the geometric means across quartiles suggests that the OH-PAH concentrations measured in a single sample are predictive of the four-year average concentrations.

1-Naphthol is a metabolite of both carbaryl and naphthalene; 2-naphthol is a metabolite of naphthalene but not carbaryl (ATSDR, 1995). To differentiate subjects for which carbaryl was likely the major source of urinary 1-naphthol, the ratio of urinary 1-naphthol to 2-naphthol was calculated and a ratio greater than two (1-naphthol/2-naphthol > 2) was used as a cut-point (Meeker et al., 2007). We determined whether there was a different distribution of covariates among girls above and below the cut-point by calculating a chi-square test for each categorical variable and t-test for urinary cotinine. In sensitivity analyses, participants with 1-naphthol/2-naphthol > 2 were excluded to focus on predictors and temporal variability of PAH exposure specifically when examining 1-naphthol.

3. Results

Participants had a mean age of 7.3 years at baseline (range 6.4–8.1) and represented diverse race/ethnic groups (black, Hispanic, Asian, and white) (Table 1). Almost 30% of girls were overweight (> 85th age- and sex-adjusted BMI percentile). The participants’ households had fairly high socioeconomic status (SES) as measured by parent/guardian education level and income.

Table 1.

Characteristics at baseline for 431 girls with urinary PAHs measurements: BCERP cohort Northern California site, 2005–2006.

	N	%
Age (years)
6.0–6.9	102	24
7.0–7.9	322	75
≥ 8.0	7	2
Child race/ethnicity
Black	96	22
Hispanic	106	25
Asian	52	12
White	177	41
BMI percentilea
< 50th	144	33
50th–85th	159	37
> 85th	128	30
Primary parent or guardian educationb
High school or less	82	19
Some college	132	31
Bachelor's degree or greater	215	50
Household income (dollars)b
< 50,000	91	21
50,000–100,000	155	37
≥ 100,000	178	42
Grilled food in last 2 daysb
Yes	87	20
No	342	80
Distance weighted traffic density (highest AADT in 150m buffer)b
Local traffic	121	28
< 1813.47	78	18
1813.47–4536.87	77	18
4536.88–9727.07	77	18
≥ 9727.08	77	18
Season of urine collection
Spring (March–May)	117	27
Summer (June–August)	166	39
Fall (September–November)	83	19
Winter (December–February)	65	15
Region of residence at baseline
Marin County	94	22
San Francisco	83	19
East Bay	254	59

^aAge- and sex-specific BMI percentile based on 2000 growth charts from CDC.

^bCounts do not total to 431 due to missing data: 2 missing parent/guardian education, 7 missing household income, 2 missing grilled food consumption, 1 missing traffic density.

Eight of the ten PAH metabolites were detected in ≥ 95% of the samples and only one metabolite (4-hydroxyphenanthrene) in < 60% (Table 2). Urinary 1-naphthol and 2-naphthol had the highest concentrations among the PAH metabolites. Participants had higher concentrations of 1-naphthol (GM 1510) ng/L, 95% confidence interval (CI) 1320–1710 ng/L) compared to concentrations among 6–11 year olds from the NHANES 2005–2006 cycle (GM 1190) ng/L, 95% CI 1010–1400 ng/L (CDC, 2017) (p-value = 0.02), but had lower than the national levels for other OH-PAHs (Table 2).

Table 2.

Geometric mean and selected percentiles of baseline urinary PAH metabolite concentrations (in ng/L) for the BCERP CA site and U.S. population.

	BCERP CA site										U.S. populationa
						Percentiles
Analyte	N	LODb	% > LOD	GM	(95% CI)	25th	50th	75th	90th	95th	N	GM	(95% CI)	50th percentile
1-naphthol	431	28	100	1510	(1320–1710)	598	1300	3160	9690	18,600	336	1190	(1010–1400)	1030
2-naphthol	431	21	100	1570	(1420–1740)	790	1520	3210	6000	8980	345	2160	(1940–2410)	2090
2-hydroxyfluorene	431	13	99.8	140	(130–150)	92.9	145	226	337	461	343	201	(184–221)	195
3-hydroxyfluorene	431	6.0–59	95.8	61.8	(57.1–66.8)	39.9	62.5	95.5	159	243	336	81.7	(73.6–90.7)	80.6
9-hydroxyfluorene	431	9.0	99.8	172	(160–186)	102	177	302	471	587	342	184	(162–209)	190
1-hydroxyphenanthrene	431	7.0–24	98.4	75.5	(70.1–81.4)	47.3	77.5	125	178	245	345	111	(101–123)	110
2-hydroxyphenanthrene	431	4.0–24	85.9	31.7	(29.6–34.1)	18.7	31.8	49.4	80.4	108	341	42.1	(38.4–46.1)	40.1
3-hydroxyphenanthrene	431	7.0–28	95.6	66.4	(61.6–71.6)	41.3	69.3	113	174	219	345	89.6	(79.6–101)	84.0
4-hydroxyphenanthrene	431	6.0–14	58.0	–c		<LOD	8.29	17.3	30.6	44.5	307	20.4	(17.7–23.4)	20.2
1-hydroxypyrene	429	6.0–25	98.1	74.2	(68.9–80.0)	46.5	78.8	121	192	241	342	103	(91.3–117)	99.8

Abbreviations: GM geometric mean, CI confidence interval.

^aNHANES 2005–2006 6–11 year olds.

^bLOD (limit of detection) values tend to be at the lower end of these ranges, but were dependent on blanks, and calculated per participant.

^cNot calculated; proportion of results below the LOD was too high to provide a valid result.

In models including only demographics variables and adjusting for urinary creatinine (Table S1), associations were seen with younger age (6–6.9 years old compared to ≥7.0 years old) for Σfluorene (proportional change in metabolite concentration, e^β, 1.19, 95% CI 1.07–1.33) and Σphenanthrene (e^β 1.13, 95% CI 1.01–1.27); with Asian race/ethnicity for 2-naphthol (e^β 1.33, 95% CI 1.03–1.72) and 1-hydroxypyrene (e^β 1.24, 95% CI 1.02–1.50); with Hispanic race/ethnicity for 2-naphthol (e^β 1.65, 95% CI 1.33–2.05); with black race/ethnicity for 1-naphthol (e^β 0.76, 95% CI 0.54–1.07) and Σphenanthrene (e^β 0.91, 95% CI 0.79–1.04); and with lower household income for 2-naphthol (e^β 1.38, 95% CI 1.09–1.74). Further adjustment, including sources of exposure, in multivariable models attenuated these associations (Table 3). There was no association between age and any PAH metabolite concentration in these multivariable models. The associations between urinary metabolite concentrations and race remained after adjustment, as did the association of lower household income with 2-naphthol (Table 3).

Table 3.

Proportional change in urinary PAHs biomarker concentrations^a (exponentiated β) by characteristics of participants in the BCERP study, CA site.

	1-naphthol		2-naphthol		Σfluorene
	eβ	95% CI	eβ	95% CI	eβ	95% CI
Race
Black	0.73	(0.51–1.04)	1.07	(0.84–1.36)	0.89	(0.78–1.03)
Hispanic	0.87	(0.63–1.20)	1.63	(1.30–2.04)	0.99	(0.87–1.12)
Asian	0.79	(0.54–1.16)	1.28	(0.98–1.67)	0.92	(0.79–1.07)
White	1	–	1	–	1	–
Household income (dollars)
<50,000	0.92	(0.64–1.33)	1.34	(1.05–1.72)	0.96	(0.83–1.11)
50,000–100,000	0.86	(0.66–1.12)	1.08	(0.90–1.29)	0.92	(0.83–1.02)
≥ 100,000	1	–	1	–	1	–
Cotinine (µg/L)
ln(cotinine), continuous	1.14	(1.03–1.26)	1.09	(1.01–1.17)	1.08	(1.03–1.12)
Grilled food in last 2 days
Yes	1.33	(1.00–1.78)	1.19	(0.98–1.45)	1.18	(1.06–1.33)
No	1	–	1	–	1	–
Distance weighted traffic density (highest AADT in 150m buffer)
Unexposed (no segment)	1	–	1	–	1	–
<1813.47	1.09	(0.78–1.53)	1.09	(0.86–1.37)	0.99	(0.87–1.14)
1813.47–4536.87	0.92	(0.65–1.32)	0.93	(0.73–1.19)	0.99	(0.86–1.14)
4536.88–9727.07	1.10	(0.77–1.57)	0.89	(0.70–1.13)	1.02	(0.89–1.18)
≥ 9727.08	1.18	(0.83–1.70)	0.88	(0.69–1.13)	1.05	(0.90–1.21)
Region of residence
Marin County	0.78	(0.57–1.06)	1.00	(0.80–1.24)	0.81	(0.71–0.92)
San Francisco	1.07	(0.79–1.48)	0.98	(0.78–1.22)	1.19	(1.05–1.35)
East Bay	1	–	1	–	1	–
Season of urine collection
Spring (Mar–May)	1.78	(1.33–2.38)	1.22	(0.99–1.49)	0.98	(0.87–1.10)
Summer (Jun–Aug)	1	–	1	–	1	–
Fall (Sep–Nov)	3.06	(2.13–4.40)	1.03	(0.80–1.32)	0.91	(0.79–1.06)
Winter (Dec–Feb)	1.90	(1.32–2.75)	0.92	(0.72–1.19)	0.92	(0.80–1.07)-
	Σ phenanthrene			1-hydroxypyrene
	eβ		95% CI	eβ		95% CI
Race
Black	0.87		(0.75–1.01)	1.05		(0.89–1.24)
Hispanic	1.00		(0.88–1.15)	1.09		(0.93–1.27)
Asian	1.04		(0.89–1.22)	1.15		(0.96–1.39)
White	1		–	1		–
Household income (dollars)
<50,000	0.91		(0.79–1.06)	0.96		(0.80–1.13)
50,000–100,000	0.90		(0.80–1.00)	0.90		(0.79–1.02)
≥ 100,000	1		–	1		–
Cotinine (µg/L)
ln(cotinine), continuous	1.04		(0.99–1.08)	1.04		(0.99–1.09)
Grilled food in last 2 days
Yes	1.16		(1.03–1.31)	1.12		(0.98–1.29)
No	1		–	1		–
Distance weighted traffic density (highest AADT in 150m buffer)
Unexposed (no segment)	1		–	1		–
<1813.47	0.97		(0.85–1.12)	0.90		(0.76–1.05)
1813.47–4536.87	0.93		(0.80–1.07)	0.91		(0.77–1.08)
4536.88–9727.07	0.95		(0.82–1.10)	0.86		(0.72–1.02)
≥ 9727.08	1.05		(0.90–1.22)	1.00		(0.84–1.18)
Region of residence
Marin County	0.82		(0.72–0.93)	0.90		(0.77–1.04)
San Francisco	1.30		(1.14–1.48)	1.24		(1.07–1.45)
East Bay	1		–	1		–
Season of urine collection
Spring (Mar–May)	1.06		(0.93–1.19)	1.10		(0.95–1.26)
Summer (Jun–Aug)	1		–	1		–
Fall (Sep–Nov)	0.95		(0.81–1.10)	0.89		(0.75–1.06)
Winter (Dec–Feb)	1.05		(0.90–1.23)	1.06		(0.89–1.27)

^aAdjusted for ln(cotinine), recent grilled food consumption, traffic metrics, region of residence, season of sample collection, age, race, income, and ln(creatinine).

After this further adjustment, there was a consistent association between PAH metabolite concentrations and urinary cotinine concentrations (e^β ranged from 1.04 to 1.14). The associations were the strongest for 1-naphthol, 2-naphthol, and Σfluorene (Table 3). Urinary cotinine concentrations in this study sample were relatively low (median 0.21 µg/L and GM 0.26 µg/L, 95% CI 0.23–0.29). By comparison, serum cotinine concentrations among 3–11 year olds in NHANES 2005–2006 had a median of 0.050 µg/L and GM 0.078 µg/L; urinary cotinine concentrations are typically 7–10 times higher so the estimated NHANES urinary concentrations are median of 0.35–0.50 µg/L and GM 0.55–0.78 µg/L. In our study sample, the range of urinary cotinine concentrations (0.025–129 µg/L) was wide due to one participant with an extreme value. Exclusion of participants with outlier cotinine values did not affect the associations between PAH metabolite concentrations and urinary cotinine concentrations (e^β range 1.04–1.14). The interquartile range (IQR) of urinary cotinine in this study sample is 0.114−0.474 µg/L; the proportional increase in OH-PAH concentrations corresponding to the cotinine IQR is 21% for 1-naphthol, 13% for 2-naphthol, 11% for Σfluorene, 5% for Σphenanthrene, and 6% for 1-hydroxypyrene.

A consistent association between PAH metabolite concentrations and recent grilled food consumption (e^β range 1.12–1.33) was observed; the 95% CIs of these associations did not exclude the null for all PAH metabolites (Table 3). There was no association found between OH-PAH concentrations and distance-weighted traffic density around the girl's residence. For all metabolites except 2-naphthol, residence in San Francisco was associated with higher concentrations, while residence in Marin was associated with lower concentrations; these associations were the strongest for Σfluorene, Σphenanthrene, and 1-hydroxypyrene. Naphthalene metabolites concentrations varied by season: 2-naphthol concentrations were somewhat higher in the spring; 1-naphthol concentrations were higher in the spring and winter and highest in the fall, compared to summer (Table 3).

Most of the PAH metabolites were correlated with each other (Table S2); 1-naphthol correlation was comparatively weak with 2-naphthol (0.34) and weak-to-moderate with the other metabolites (0.33–0.45), suggesting other exposure sources for 1-naphthol. We excluded participants with 1-naphthol/2-naphthol > 2 (suggesting likely carbaryl exposure (Meeker et al., 2007), 24.6% of participants) and obtained similar results for 1-naphthol predictors as in the model with all participants: 1-naphthol concentrations were positively associated with urinary cotinine concentrations (e^β 1.12, 95% CI 1.03–1.22) and recent grilled food consumption (e^β 1.19, 95% CI 0.94–1.51); neither distance-weighted traffic density around the girl's residence nor region of residence were associated; concentrations were still higher in the spring (e^β 1.28, 95% CI 1.00–1.63) and winter (e^β 1.52, 95% CI 1.12–2.07), and while still elevated in the fall, to a lesser degree than the full sample (e^β 1.58, 95% CI 1.14–2.19). Participants with 1-naphthol/2-naphthol > 2 were more likely to have a fall sample collection, while those with a ratio at or less than two were more likely to have a summer sample collection (Table S3). The excluded girls were more likely to be white, of higher SES and somewhat more likely to live in San Francisco than other girls (Table S3).

Among the subset with repeated measures (N = 100), the within-person variance was larger than the between person variance (Table 4). 1-Naphthol showed substantial within-person variability, whereas the other OH-PAH metabolites showed comparatively less variability. The ICCs indicated fair (0.20–0.39) to moderate (0.40–0.59) reproducibility for seven metabolites (i.e., 2-naphthol, 2-hydroxyfluorene, 9-hydroxyfluorene, 1-hydroxyphenanthrene, 2-hydroxyphenanthrene, 3-hydroxyphenanthrene, 1-hydroxypyrene) and both sums (i.e., Σfluorene, Σphenanthrene). The reproducibility was poor for 1-naphthol (ICC = 0.12) and 3-hydroxyfluorene (ICC = 0.18). The Spearman correlation coefficients were weak for 1-naphthol, but moderate (> 0.4) for at least two of the three time intervals for four metabolites (i.e., 2-naphthol, 2-hydroxyfluorene, 1-hydroxyphenanthrene, 3-hydroxyphenanthrene) and Σphenanthrene. Except for 1-naphthol, when the four-year average concentrations were categorized into quartiles according to the quartile distributions at each time point, the geometric mean increased monotonically from lowest to highest quartile, thus supporting the use of the concentrations from a single spot urine sample to correctly categorize a subject's average, multi-year exposure into high to low quartiles (Table 5). When we excluded 1-naphthol /2-naphthol > 2 samples (23% of baseline, 29% of year 2, 17% of year 4), we obtained similar results for the 1-naphthol ICC and Spearman correlations coefficients. Excluding participants with any sample that had a 1-naphthol /2-naphthol > 2, the remaining 54 girls had a non-monotonic increase in the 1-naphthol four-year geometric means across the quartiles.

Table 4.

Components of variance, intraclass correlation coefficients (ICC), and spearman correlation coefficients for creatinine adjusted urinary OH-PAHs between samples collected at baseline, one and three years after enrollment for 100 girls in the BCERP study, CA site.

	Variance component		ICC	Spearman correlation coefficients (p-value)
	Between-child	Within-child	(95% CI)	Baseline & Year 2	Baseline & Year 4	Year 2 & Year 4
1-naphthol	0.15	1.07	0.12	0.07	0.17	0.13
			(0.04, 0.31)	(0.52)	(0.09)	(0.18)
2-naphthol	0.27	0.42	0.39	0.43	0.52	0.37
			(0.28, 0.52)	(<0.0001)	(<0.0001)	(<0.001)
2-hydroxyfluorene	0.11	0.18	0.37	0.33	0.48	0.44
			(0.26, 0.50)	(<0.001)	(<0.0001)	(<0.0001)
3-hydroxyfluorene	0.07	0.32	0.18	0.15	0.28	0.28
			(0.08, 0.34)	(0.13)	(<0.01)	(<0.01)
9-hydroxyfluorene	0.19	0.27	0.41	0.32	0.37	0.42
			(0.29, 0.53)	(<0.01)	(<0.001)	(<0.0001)
Σfluorene	0.11	0.20	0.35	0.25	0.41	0.39
			(0.24, 0.49)	(0.01)	(<0.0001)	(<0.0001)
1-hydroxyphenanthrene	0.14	0.17	0.45	0.49	0.47	0.44
			(0.34, 0.57)	(<0.0001)	(<0.0001)	(<0.0001)
2-hydroxyphenanthrene	0.11	0.22	0.33	0.34	0.19	0.50
			(0.22, 0.47)	(<0.001)	(0.07)	(<0.0001)
3-hydroxyphenanthrene	0.17	0.17	0.49	0.45	0.41	0.53
			(0.38, 0.60)	(<0.0001)	(<0.0001)	(<0.0001)
Σphenanthrene	0.13	0.15	0.47	0.43	0.37	0.50
			(0.35, 0.58)	(<0.0001)	(<0.001)	(<0.0001)
1-hydroxypyrene	0.09	0.26	0.26	0.41	0.33	0.38
			(0.16, 0.41)	(<0.0001)	(<0.001)	(<0.001)

Table 5.

Four-year geometric means of urinary PAH metabolite concentrations (ng/g creatinine) within surrogate sample determined exposure quartiles for 100 girls in the BCERP study, CA site.

	Four-year geometric mean of each single year determined quartile
	Q1	Q2	Q3	Q4
1-naphthol
baseline	1150	1140	1700	2940
year 2	907	1384	1720	3050
year 4	982	1510	1420	3120
2-naphthol
baseline	880	1480	2150	3380
year 2	945	1360	2170	3400
year 4	997	1460	1850	3530
Σfluorene
baseline	322	406	497	678
year 2	331	390	436	783
year 4	319	392	491	716
Σphenanthrene
baseline	165	209	218	389
year 2	162	201	220	408
year 4	161	199	242	379
1-hydroxypyrene
baseline	84.8	107	138	190
year 2	81.9	116	129	195
year 4	85.1	116	130	186

Bold text denotes non-monotonic increase in four-year geometric mean across quartiles.

Abbreviations: Q quartile.

4. Discussion

The aim of this study was to determine the distribution of PAH metabolite concentrations in a peri-pubertal sample of Californian girls to identify predictors and potential sources of exposure. Our results show that these compounds were detected in a large proportion of young girls, with 1-naphthol at somewhat higher geometric mean concentrations than seen nationally (CDC, 2017). The data do not suggest a generally high level of PAH exposure among girls living in the San Francisco Bay area; urinary concentrations of PAH metabolites were similar to (Cavanagh et al., 2007; Freire et al., 2009; Health Canada, 2015; Wilhelm et al., 2008) or lower than (Alghamdi et al., 2015; Fiala et al., 2001; Hemat et al., 2012; Jung et al., 2014; Kang et al., 2002; Martinez-Salinas et al., 2010; Morgan et al., 2015; Mucha et al., 2006; Ochoa-Martinez et al., 2016; Thai et al., 2016; van Wijnen et al., 1996; Yoon et al., 2012) concentrations detected in children in other studies. Nevertheless, even low-level environmental PAH exposure is a public health concern because of the observed estrogenic properties of PAHs (Li et al., 2012a) and OH-PAHs (Li et al., 2012a; Schultz and Sinks, 2002; Sievers et al., 2013; Wenger et al., 2009) and therefore their potential to affect development during the susceptible pubertal window.

The girls were exposed to a mixture of PAHs in their environment as evident from the high-detection rates of multiple OH-PAHs. Studies that solely measure 1-hydroxypyrene may miss key demographic differences, potential sources, and temporal variability across metabolites. Inhalation has been found to be the exposure source for naphthalene metabolites (ATSDR, 2005; Li et al., 2010a), while fluorene, phenanthrene, and pyrene metabolites also suggest dietary exposure (Li et al., 2012b). Metabolism of PAHs may differ between exposure sources; Heudorf and Angerer (2001) found that smoking induces increased metabolism of phenanthrene into 3-hydroxyphenanthrene and 4-hydroxyphenanthrene.

In general, concentrations of OH-PAHs varied by child's race/ethnicity. 2-Naphthol concentrations were the highest among Hispanic girls, and there was some evidence that 2-naphthol and 1-hydroxypyrene concentrations were higher among Asian girls, while 1-naphthol and Σphenanthrene concentrations were lower among black girls, compared to white girls. We did not find that OH-PAH concentrations varied substantially by SES, nor did this explain the race differences. The exception was 2-naphthol; concentrations were highest among those in the lowest household income group (< $50,000/year). The initial observation of variation by age was attenuated after adjustment for sources of exposure. Our findings for Hispanic girls are consistent with NHANES data showing that Mexican-American children had higher urinary 2-naphthol concentrations than non-Hispanic white children, but had lower concentrations of other OH-PAHs (Perla et al., 2015).

Cotinine concentrations in this study sample were relatively low compared to NHANES and yet cotinine was predictive of OH-PAH concentrations. It is not surprising that the associations were strongest for 1-naphthol, 2-naphthol, and Σfluorene as naphthalene and fluorene are the two most abundant PAHs in cigarette smoke (Ding et al., 2005). A PAH biomonitoring study comparing adult smokers and non-smokers found naphthalene and fluorene metabolites to have the highest probability of predicting smokers (St Helen et al., 2012). In contrast, in our study, smoke exposure is most likely due to environmental tobacco smoke (ETS) as these girls were only 6–8 years old when cotinine was measured. An association between low cotinine concentrations indicating ETS exposure and PAH metabolites is not consistently found in the literature. In a study of 101 children under 6 years old, 1-hydroxypyrene was strongly correlated with urinary cotinine, while the phenanthrene metabolites were not; there was some evidence indicating that ETS (assessed via questionnaire) was positively associated with PAH metabolite concentrations (Heudorf et al., 2001). Other studies in adults evaluating exposure to PAHs (via air sampling of PAHs and urine concentrations of PAH metabolites) did not find associations between reported ETS exposure and 2-naphthol or phenanthrene metabolites (Aquilina et al., 2010), nor 1-hydroxypyrene (Scherer et al., 2000). An analysis of 1999–2002 NHANES data, which classified ≥6 year old participants by serum cotinine concentrations into the categories of no exposure, involuntary exposure, and active smoking, found that involuntary smoke exposure was strongly associated with urinary concentrations of fluorene, phenanthrene, and pyrene metabolites (Suwan-ampai et al., 2009). The differences in results across studies, including our own, may be due to other background sources of PAHs or levels of ETS, especially for children compared to adults, or measurement tools used to define ETS exposure (e.g., questionnaire, air sampling, biomonitoring).

Recent grilled food consumption was also associated with OH-PAH concentrations in our study. Diet, particularly the consumption of grilled meat or vegetables, has been reported to be the main source of PAHs for non-smokers (ATSDR, 1995; Yebra-Pimentel et al., 2015). However, the association was not statistically significant for all of the OH-PAHs, which may be due to individual metabolic variability. Li et al. (2012b) reported that after dietary exposure, on average, 58–79% of OH-PAH metabolites (same metabolites measured in this study) were excreted in the first 12 h and concentrations returned to background within 24–48 h, so our consumption questions asking about the last 2 days should be relatively specific.

Vehicle exhaust has been reported to be a source of non-occupational PAH exposure, particularly among people living in areas with high traffic density (Mordukhovich et al., 2016; Tuntawiroon et al., 2007). However, an association with traffic density was not observed in this study sample. We determined the traffic density within a 150m buffer around the residential address, because previous studies have found that 80–90% of ambient air pollutants decay between 150 and 200m (English et al., 1999). It is possible that the 150m buffer was too narrow to capture the traffic density the child is typically exposed to (e.g., a child spends a majority of the day in other locations such as school, typically travels more than 150m from the home). Also, the traffic density was calculated from annual AADT data from 2004; it would not account for recent traffic changes in the girl's neighborhood prior to sample collection.

We found that region of the San Francisco Bay area (i.e., San Francisco Peninsula, North Bay, and East Bay) was associated with OHPAHs concentrations after adjusting for traffic density, demographics, and other exposures. This regional variability is consistent with a PAH dust study conducted in 2001–2007 that reported higher PAH concentrations in the dust of metropolitan San Francisco Bay area households than northern San Francisco Bay area households (Whitehead et al., 2013). In that study,Whitehead et al. (2013) found that the metropolitan San Francisco Bay area region had the densest population, with many residential and commercial buildings, most traffic emissions, and the oldest homes. The region, rather than the traffic density as measured, could be more accurately capturing the urban status of the residential address or could be capturing regional variability in other factors that can affect PAH exposure (e.g., ambient air pollution, presence of attached garages, industrial areas).

Previous studies have found seasonality of PAH biomarker concentrations (Cavanagh et al., 2007; Jung et al., 2014; Morgan et al., 2015), specifically higher PAHs associated with the months that indoor space heating (e.g., furnace, fireplace) would be used (Cavanagh et al., 2007; Jung et al., 2014), while other studies have found no differences by season (Peters et al., 2017). We found 1-naphthol concentrations were lowest in summer, and 2-naphthol concentrations were somewhat higher in the spring. Seasonal pattern could vary from other studies, because the Bay Area has a temperate climate, or because we controlled for some of the other factors for which season may be a proxy (e.g., traffic, grilled food consumption).

While the other OH-PAHs were correlated with each other, 1-naphthol was not, including with 2-naphthol. 1-Naphthol is a metabolite of both carbaryl and naphthalene. Carbaryl is quickly metabolized into 1-naphthol; 1-naphthol accounts for over 85% of all carbaryl urinary metabolites (DPR, 2014b; Maroni et al., 2000). Carbaryl or 1-naphthyl methylcarbamate is a widely used insecticide. In 2005 and 2006, over 400,000 pounds of carbaryl were sold in California each year, of which about 40% was applied agriculturally to crops (e.g., fruits, vegetables, nuts) (DPR, 2014a, 2014b). Peak application times vary by crop (e.g., tomatoes high-use season March-May, citrus highuse season May-October) (DPR, 2014b). Nationally, it was the most commonly used conventional insecticide active ingredient in the home and garden market in 2005 (US EPA, 2011). However, exposure to carbaryl is not exclusively from the diet or garden;Li et al. (2012b) reported elevated 1-naphthol concentrations in a person who had exercised in the morning at a gym that was freshly sprayed with a cleaner or pesticide. We identified 24.6% of participants with 1-naphthol/2-naphthol > 2 (suggestive of likely carbaryl exposure (Meeker et al., 2007). This is similar to the proportion reported in other U.S. study samples (Castorina et al., 2010; Li et al., 2012b, 2016) and higher than the proportion reported in other non-U.S. study samples (Wheeler et al., 2014); to our knowledge, this is the first study to document this finding in a sample of children. A higher proportion of girls with likely carbaryl exposure were white, higher SES as defined by parent education or household income, and lived in San Francisco. The excluded girls had similar exposure to combustion by-products (e.g., cotinine concentrations, grilled food consumption, traffic density), which addresses the concern that the 1-naphthol/2-naphthol > 2 cut-off excludes those with low PAH exposure (i.e., relatively smaller exposure to carbaryl could change the metabolite ratio proportionately more than when overall PAH exposure is higher). When the 1-naphthol analyses were re-run without these girls in order to focus on naphthalene exposure, the results were confirmed except the association with fall season was substantially attenuated.

Among the subset of subjects with repeated samples, we found moderate reproducibility over the four-year period examined for most PAH metabolites even with substantial within-person variability. The ICCs were fair to moderate for seven of the nine metabolites and for both sums. Since the samples were collected at least one year apart, these values reflect day-to-day variability as well as long-term variability (e.g., changes in diet, activity). The ability of a single spot urine sample to determine a subject's multi-year average exposure varied across the PAH metabolites. For 1-naphthol, a single urine sample was the least predictive and excluding participants with 1-naphthol/2-naphthol > 2 did not improve its predictive ability. We expect that health studies with a single urinary 1-naphthol measurement would have non-differential exposure misclassification, which would tend to bias results towards the null and limit the ability to detect an association between the outcome and 1-naphthol. For the other metabolites, concentrations from a single spot urine sample were able to accurately rank exposure into concentration quartiles consistent with the multi-year average PAH metabolite concentration quartiles. This implies accurate exposure assessment with a single sample to define multi-year exposure timeframe. These data confirm ubiquitous exposure to PAHs (CDC, 2017); because PAH exposures arise from common, daily sources, there is reasonable reproducibility of quickly metabolized, nonpersistent compounds over a four-year timeframe. Chronic, long-term exposures resulting in reasonably stable concentrations of exposure biomarkers have been documented for other nonpersistent environmental chemicals as well, e.g. perchlorate, phenols, phthalates (Mervish et al., 2016; Teitelbaum et al., 2008; Townsend et al., 2013).

The study has several limitations. We were not able to account for some variables reflecting potential exposure sources (e.g., attached garage) or other risk factors that might affect metabolism (e.g., stress, genetic variability). Second, we collected spot urine samples and OHPAHs have short half-lives so these results may not be representative of long-term PAH exposure. However, except for 1-naphthol, our findings for repeatability over a four-year timeframe ranged from poor to moderate agreement (ICCs between 0.18 and 0.49), depending on the metabolite, and the analysis using surrogate rankings had excellent agreement for OH-PAH quartiles, suggesting consistent exposure over time. Lastly, the results might not be generalizable to the general population, as participants’ households were fairly high SES and families had agreed to participate in a multi-year longitudinal study. Also, results for girls residing in the Bay Area may not be generalizable to other parts of California, the United States, or abroad.

In conclusion, our results indicate high detection rates of OH-PAHs in a diverse sample of pre- and peri-pubertal girls. The paper confirms exposure to even relatively low levels of environmental tobacco smoke and recent grilled food consumption as the main sources of PAHs, but does not find an association with traffic density, as measured. Seasonal patterns among California girls vary from studies conducted elsewhere. The smoke exposure, diet, and traffic density results were consistent across the PAH metabolites and after adjusting for known confounders. Controlling for these predictors, differences still existed by race for specific PAH metabolites and by income for 2-naphthol.

Abbreviations

AADT

average annual daily traffic

BCERP

Breast Cancer and the Environment Research Program

BMI

body mass index

CDC

Centers for Disease Control and Prevention

CEHTP

California Environmental Health Tracking Program

CI

confidence interval

ETS

environmental tobacco smoke

GM

geometric means

ICC

intraclass correlation coefficients

IQR

interquartile range

IRB

institutional review board

KPNC

Kaiser Permanente Northern California

LOD

limit of detection

NHANES

National Health and Nutrition Examination Survey

OH-PAHs

monohydroxy-polycyclic aromatic hydrocarbons

PAHs

polycyclic aromatic hydrocarbons

SES

socioeconomic status