Environmental styrene exposure and neurologic symptoms in U.S. Gulf coast residents

Abstract

Background:

Styrene is an established neurotoxicant at occupational levels, but effects at levels relevant to the general population have not been studied. We examined the neurologic effects of environmental styrene exposure among U.S. Gulf coast residents.

Methods:

We used National Air Toxics Assessment 2011 estimates of ambient styrene concentrations to assign exposure levels for 21,962 non-diabetic Gulf state residents, and additionally measured blood styrene concentration in a subset of participants (n=874). Neurologic symptoms, as well as detailed covariate information, were ascertained via telephone interview. We used log-binomial regression to estimate prevalence ratios (PR) and 95% confidence intervals (95% CI) for cross-sectional associations between both ambient and blood styrene levels and self-reported neurologic symptoms. We estimated associations independently for ten unique symptoms, as well as for the presence of any neurologic, central nervous system (CNS), or peripheral nervous system (PNS) symptoms. We also examined heterogeneity of associations with estimated ambient styrene levels by race and sex.

Results:

One-third of participants reported at least one neurologic symptom. The highest quartile of estimated ambient styrene was associated with one or more neurologic (PR, 1.12; 95% CI: 1.07,1.18), CNS (PR, 1.17; 95% CI: 1.11,1.25), and PNS (PR, 1.16; 95% CI: 1.09,1.25) symptom. Results were less consistent for biomarker analyses, but blood styrene level was suggestively associated with nausea (PR, 1.78; 95% CI: 1.04, 3.03). In stratified analyses, we observed the strongest effects among non-White participants.

Conclusions:

Increasing estimated ambient styrene concentration was consistently associated with increased prevalence of neurologic symptoms. Associations between blood styrene levels and some neurologic symptoms were suggestive. Environmental styrene exposure levels may be sufficient to elicit symptomatic neurotoxic effects.

INTRODUCTION

Styrene is a hydrocarbon used in the production of plastics, fiberglass laminates, rubber, and resins found in consumer products and commercial and residential building materials. Manufactured styrene products include insulation, fiberglass boats, automotive parts, car tires, Styrofoam, and plastic drinking glasses ¹. Styrene is an established neurotoxicant at occupational levels ^1–3, but has not been studied at environmental levels experienced outside of occupational contexts. Epidemiologic studies to date have focused on highly exposed workers, whose average blood concentrations are 25 times higher than those of the general population ^4–11.

Inhalation of tobacco smoke, off-gassing of building materials and consumer products, and vehicle and industrial emissions accounts for over 90% of styrene exposure in the general population ^12,13,3,14. Exposure via dermal contact is typically limited to occupational settings, and internal dose due to ingestion from food and water is considered negligible. Among smokers, cigarettes are considered the dominant source of styrene exposure, and smoking is the single most important individual predictor of human exposure to styrene ^1,15–21. Styrene is released into the air from automobile exhaust, cigarette smoke, photocopiers and printers, and industries using or manufacturing styrene. The Gulf region is home to a prolific petrochemical industry, many styrene-emitting industries, and over half of all U.S. styrene production ^1,22, potentially exposing Gulf residents to a high intensity of environmental styrene emissions.

Styrene is commonly detected in urban air, near industrial sites and landfills, and in high traffic areas, although typically at levels substantially lower than in occupational settings. Rural and suburban air generally contains lower concentrations of styrene than urban air ²³. In the U.S. general population, daily exposure to styrene in air is estimated to be 18–54 μg/person ¹, and indoor air usually contains higher levels of styrene than outdoor air. Because the half-life of styrene in blood is approximately 13 hours ², blood styrene measurements reflect recent exposure.

The Agency for Toxic Substances and Disease Registry (ATSDR) has identified the central nervous system (CNS) as the primary target for styrene toxicity, with less marked effects in the peripheral nervous system (PNS) ^12,24. Like many other volatile organic compounds, styrene monomer is a CNS depressant with anesthesia-like properties ^3,25. Acute solvent-induced neurotoxicity, including that caused by styrene, is characterized by symptoms of acute intoxication, commonly described as a feeling of drunkenness. Long term exposures at levels found in occupational settings have been associated with chronic adverse neurotoxic effects. Occupational studies demonstrate styrene-induced neurotoxicity, from both acute and chronic inhaled exposure among highly-exposed workers. Symptoms include feeling “drunk” and tiredness ²⁶, impaired vision ^24,27, vestibular dysfunction ⁹, headaches ²⁸, delayed reaction time ^29,30, impaired attention and memory ⁶, hearing deficits ³¹, diminished nerve conduction velocity ^6,32–35, and abnormal EEG results ^35,36. Similar effects have been observed at lower occupational airborne exposure levels, ranging from 10–30 ppm, in most ^{7,24,28–30,37–40}, though not all ^10,41, studies.

The human health effects of chronic styrene exposure at typical environmental levels remain largely unknown ¹⁸. We set out to assess the associations between two metrics of styrene exposure - estimated ambient concentrations and measured blood levels - and self-reported neurologic symptoms among Gulf states residents. Quantifying the association between environmental styrene exposure and highly sensitive, but non-specific, neurologic symptoms may lend insight into early manifestations of environmentally induced neurotoxicity due to chronic exposures at levels insufficient to cause clinically apparent toxicity.

METHODS

Study Design and Participants

We used data from the Gulf Long-term Follow-up Study (GuLF STUDY), a prospective cohort of adults (ages 21 and older) who participated in oil spill response activities and others who received safety training, but were not hired, following the 2010 Deepwater Horizon disaster. A detailed description of this study is available elsewhere ⁴². Of the 25,848 English- or Spanish-speaking GuLF STUDY participants living in the Gulf region (Alabama, Florida, Louisiana, Mississippi, and Texas) at enrollment, 24,903 reported addresses that were successfully geocoded to a 2010 U.S. Census tract. From this sample of participants with known residential locations, we excluded participants with any missing neurologic symptom information (n=304), missing demographic characteristics (n=573), and missing covariate information (n=201), leaving 23,825 eligible participants. Because autonomic and peripheral neuropathy are known complications of diabetes ⁴³, we further restricted the study sample to participants with no self-reported physician diagnosis of diabetes (exclude 1,825 diabetics and 38 with missing diagnosis information), resulting in a final analytic sample of 21,962.

Approximately 2–3 years after the oil spill (May 2012-July 2013), a subset of GuLF STUDY participants living in the Gulf region (N=1,055) were enrolled in the Chemical Biomonitoring Study (CBS) ⁴⁴. CBS participants provided an extra blood sample for measuring styrene and other compounds and completed a questionnaire about usual and recent exposures and exposure opportunities.

Ultimately, 994 participants provided blood samples sufficient for quantification of styrene levels. Of those, we excluded 20 participants missing neurologic symptom information, nine with incomplete demographic characteristics, and four individuals missing other covariate information, leaving 961 eligible participants. We then excluded known diabetics (n=86 of the 1,825 diabetics from the parent study) or those missing diagnosis information (n=1), for a final analytic sample of 874.

Participants provided written informed consent, and the Institutional Review Board of the National Institute of Environmental Health Sciences approved this study.

National Air Toxics Assessment

The United States Environmental Protection Agency (EPA) 2011 National-scale Air Toxics Assessment (NATA) ⁴⁵ evaluates 180 air toxics across the United States using emissions inventories, dispersion modeling, photochemical modeling, exposure modeling, and toxicity analyses. NATA generates annual average ambient air toxic concentrations (µg/m³) for each U.S. census tract. We employed NATA styrene estimates as indicators of typical environmental exposure by mapping each participant’s geocoded home location to a corresponding 2010 U.S. census tract. Geocodes were based on self-reported home address at enrollment. The 2011 NATA estimated annual average ambient styrene concentration corresponding to an individual’s home census tract was applied as a surrogate of usual ambient styrene exposure for each cohort member residing in the Gulf region.

Blood styrene measurement

Blood collection tubes containing potassium oxalate and sodium fluoride anticoagulant were used to collect 10 mL of blood for styrene measurement. Tubes and stoppers were pre-treated by the Centers for Disease Control and Prevention (CDC) laboratory to remove VOC residues to minimize pre-collection contamination ^46,47. Samples were stored in a 4°C refrigerator prior to being shipped overnight on cold packs in biweekly batches to the Division of Laboratory Sciences, National Center for Environmental Health, CDC in Atlanta, Georgia, for analysis of VOCs. Analysis of styrene followed standard CDC procedures, using equilibrium headspace solid-phase micro-extraction with benchtop gas chromatography/mass spectrometry ^48,49. The laboratory provided actual measured values below the limit of detection (LOD) (0.03 ng/mL).

Neurologic symptoms

Health information, including questions about neurologic symptoms, was collected at enrollment via Computer Assisted Telephone Interview (CATI) during the baseline interview. All participants were asked to report how often they experienced dizziness, lightheadedness, vision impairment, numbness, tingling, stumbling while walking, headaches, and fatigue during the preceding 30 days. Three additional episodic outcomes, i.e., seizures, vomiting, and insomnia, were added to the questionnaire after data collection was underway. These latter outcomes were not combined with other symptoms for evaluation of “any” neurologic symptom or symptom clusters because their enumeration in the study population is incomplete (i.e., those people who completed the enrollment interview prior to the inclusion of questions about seizures, vomiting, and insomnia did not have the opportunity to report them). We did, however, analyze these additional symptoms individually among the subgroup of participants who had the opportunity to report them.

Frequency of most symptoms was reported as: all of the time, most of the time, sometimes, rarely, or never. Symptoms were classified as a binary indicator of the ‘presence’ (all or most of the time) or ‘absence’ (sometimes, rarely, or never) of occurrence. Seizures, vomiting and insomnia were reported as: every day, several times a week, once a week, rarely, or never. We applied a lower threshold for the reporting of episodic symptoms because the relative frequency of their occurrence is expected to be lower, as compared with potentially ongoing symptoms. Accordingly, the presence of vomiting included responses of once a week or more, and the presence of seizures included reporting a seizure at any point during the 30 days.

Statistical analysis

We used multivariate log-binomial regression to estimate prevalence ratios and corresponding 95% confidence intervals (PR, 95% CI) for the cross-sectional associations between modeled and measured styrene levels and neurologic symptom prevalence. We evaluated associations between (i) ambient styrene concentration estimates (µg/m³) and neurologic symptoms among all eligible study participants residing in the Gulf region (N=21,962), and (ii) blood styrene concentration (ng/mL) and neurologic symptoms among CBS participants (n=874). We also examined the distributions of NATA estimated ambient concentrations and log-transformed blood styrene levels to determine relevant rank-based exposure metrics.

We analyzed symptom clusters as primary outcomes. Based on results of a latent class analysis of symptom correlations among all reported symptoms (data not shown), we identified two neurologic clusters (i.e., CNS and PNS). The CNS cluster included dizziness, headache, nausea, sweating, and palpitations. The PNS cluster included tingling and numbness in the extremities, blurred vision, and stumbling while walking. For main analyses, we separately examined associations between styrene level and the presence of any neurologic symptom, any CNS symptom, more than one CNS symptom, any PNS symptom, and more than one PNS symptom. We included sweating and palpitations in our identification of CNS symptoms, based on results of the latent class analysis. We did not, however analyze these symptoms as individual neurologic outcomes in any other analyses because of their non-specific nature and lack of precedent in neurotoxicity literature. We further conducted sensitivity analyses excluding sweating and palpitations from the CNS symptom cluster. In secondary analyses, we examined associations between styrene levels and each individual neurological symptom in a separate model.

All models were adjusted for sex (male, female), age (<30, 30–45, >45), season (spring, summer, fall, winter), race (white, black, other), educational attainment (< high school, high school graduate or equivalent, some college, college graduate or more), employment status (currently working, not working), alcohol drinking (current, former/never), and self-reported smoking (current, former/never). In models assessing associations with blood styrene, we accounted for the duration of time (days) between enrollment (symptom ascertainment) and the date of blood styrene collection two ways. In primary analyses, we adjusted for the duration, and in sensitivity analyses, we restricted to participants within the median lag between enrollment and blood styrene collection (100 days). For statistical analyses, we used all measured blood styrene values, including the measured values below the LOD ⁵⁰. Covariates were selected based on directed acyclic graph analysis ⁵¹ of the theoretical relationship between styrene exposure and neurologic symptoms. Covariate information was obtained during the enrollment interview.

We conducted several sensitivity analyses, examining the impact of spatial clustering, smoking, co-exposures, oil spill cleanup, depression, chronic conditions, amount of overall symptom reporting (i.e., number of symptoms reported), sex, race, and socioeconomic status on observed associations. To address concern over possible effects due to spatial clustering in the study population, we used generalized estimating equations to specify a random effect of census tract (the unit of aggregation for ambient styrene estimates generated by NATA) for all ambient styrene analyses.

Given that smoking is a principal source of non-occupational styrene exposure, we evaluated smoking by adjusting for current, former, and never smokers distinctly, as well as restricting analyses to never smokers.

For co-exposure analyses, we additionally adjusted for benzene and toluene, in separate models. Ambient benzene and toluene estimates were abstracted from NATA 2011, and blood benzene and toluene levels were measured on the same panel as styrene. In ambient analyses, we additionally adjusted for ambient particulate matter (PM 2.5). Annual census tract estimates of PM 2.5 were obtained from the EPA fused Downscaler Model 2011 ⁵².

Although oil spill cleanup is not a known source of styrene exposure, we conducted sensitivity analyses assessing effects of oil spill cleanup due to its importance in defining this study population. To that end, we conducted stratified analyses of cleanup workers and non-workers separately and evaluated the statistical interaction between styrene levels and cleanup worker status.

Depression was defined based on self-reported physician diagnosis at enrollment, and we conducted stratified analyses comparing associations between participants who did and did not report depression. In addition to evaluating depression, we examined the influence of other chronic conditions on observed associations. We simultaneously adjusted for any reported cancer, emphysema, and asthma, as well as separately evaluating the impact of adjusting for diabetes instead of excluding participants reporting a diabetes diagnosis. Participants were asked about frequency of 18 non-neurologic symptoms in addition to those included here as indicative of potential neurological effects. We tried to account for symptom over-reporting by making exclusions for different thresholds of number of symptoms reported, adjusting for total number of symptoms reported, excluding participants who reported excessive recent hair loss, which is not believed to be related to any of the exposures under investigation, and restricting analyses to participants reporting only neurologic symptoms.

We evaluated effect modification by both sex and race using interaction terms between exposure and the modifier of interest, as well as running stratified models for the subgroups identified. To address socioeconomic status, we adjusted for income and self-reported concerns over affording housing and food.

We also completed analyses specifying different exposure contrasts for both ambient and blood styrene levels and repeated all main analyses restricting to the subgroup of participants who had the opportunity to report on the three symptoms (seizures, vomiting, and insomnia) that were added later.

All statistical analyses were conducted in SAS 9.4 (Cary, NC, USA).

RESULTS

This cohort is predominantly male (80.6%), younger than 45 years (58.8%), and overweight (40.4%) or obese (31.5%) (Table 1A). Approximately one-third of participants are current smokers, and 75.6% completed at least one day of oil spill response or cleanup work. Participants who provided blood specimens for CBS were less likely to be employed (52.9% vs. 62.7%) and more likely to have worked on oil spill response (84.9% vs. 75.6%), compared to the full cohort of Gulf State residents (Table 1B). The CBS also has a higher proportion of nonwhite participants than the overall sample (49.8% compared to 37.6%), and fewer college graduates (12.0% compared to 18.5%). African Americans are disproportionately likely to live in areas with higher estimated ambient styrene levels, representing 18.0% of the lowest quartile and 42.6% of the highest quartile population (Table 1A). The distribution of demographic characteristics is similar between participants who were excluded due to incomplete covariate information and participants who were included in the modeled samples.

Table 1A.

Demographic characteristics of participants living in the Gulf States (N=21,962) by estimated ambient styrene levels.

				Estimated styrene, quartiles (%)
Characteristic		N	%	Q1	Q2	Q3	Q4
Sex	Female	4,259	19.4	17.6	19.0	20.0	21.1
	Male	17,703	80.6	82.4	81.0	80.0	78.9
Age, years	< 30	4,685	21.3	20.8	19.8	21.4	23.3
	30 – 45	8,241	37.5	35.3	36.1	38.2	40.7
	> 45	9,036	41.1	43.9	44.2	40.4	36.1
Season of enrollment	Spring	6,104	27.8	28.0	26.6	27.7	28.8
	Summer	5,329	24.3	24.9	25.1	23.5	23.5
	Fall	4,968	22.6	21.9	23.8	23.2	21.6
	Winter	5,561	25.3	25.1	24.5	25.6	26.1
Race	White	13,723	62.5	70.9	70.1	60.6	48.2
	Black	6,051	27.6	18.0	20.7	29.2	42.6
	Other	2,188	10.0	11.2	9.2	10.2	9.2
Education	< High school	4,072	18.5	23.6	17.1	17.0	16.4
	High school graduate	7,285	33.2	36.9	31.6	31.1	33.0
	Some college	6,554	29.8	26.4	31.7	31.2	30.1
	≥ College graduate	4,051	18.5	13.1	19.5	20.7	20.6
Work status	Employed	13,780	62.7	63.9	63.5	63.0	60.5
	Unemployed	8,182	37.3	36.1	36.5	37.0	39.5
Current drinker	Yes	16,441	74.9	72.3	76.1	76.1	75.0
	No	5,521	25.1	27.7	23.9	23.9	25.0
Current smoker	Yes	7,514	34.2	36.8	34.1	34.1	31.7
	No	14,448	65.8	63.2	65.9	65.9	68.3
Body Mass Index, kg/m2	≤ Normal (< 25)	6,173	28.1	26.8	28.7	28.9	28.0
	Overweight (25 – <30)	8,877	40.4	39.5	41.7	40.5	40.0
	Obese (≥ 30)	6,912	31.5	33.7	29.5	30.6	31.9
Oil spill response worka	≥ 1 day	16,599	75.6	71.9	75.1	77.1	78.3
	None	5,363	24.4	28.1	24.9	22.9	21.7

^aParticipants reported working on Deepwater Horizon oil spill cleanup for at least one day in 2010–2011.

Table 1B.

Demographic characteristics of participants in the Chemical Biomonitoring Study (CBS, N=874) by measured blood styrene levels.

				Blood styrene (%)
Characteristic		N	%	≤ Median	> Median
Sex	Female	220	25.2	27.2	23.2
	Male	654	74.8	72.8	76.8
Age, years	< 30	194	22.2	23.3	21.1
	30 – 45	343	39.2	35.7	42.8
	> 45	337	38.6	41.0	36.1
Season of enrollment	Spring	174	19.9	20.4	19.4
	Summer	198	22.7	26.9	18.3
	Fall	185	21.2	20.4	22.0
	Winter	317	36.3	32.3	40.3
Race	White	439	50.2	51.8	48.6
	Black	368	42.1	38.5	45.8
	Other	67	7.7	9.7	5.6
Education	< High school	179	20.5	17.9	23.2
	High school graduate	340	38.9	36.4	41.4
	Some college	250	28.6	29.4	27.8
	≥ College graduate	105	12.0	16.3	7.6
Work status	Employed	462	52.9	58.4	47.2
	Unemployed	412	47.1	41.6	52.8
Current drinker	Yes	598	68.4	67.4	69.4
	No	276	31.6	32.6	30.6
Current smoker	Yes	261	29.9	12.7	47.4
	No	613	70.1	87.3	52.6
Body Mass Index, kg/m2	≤ Normal (< 25)	249	28.5	22.4	34.7
	Overweight (25 – < 30)	333	38.1	39.6	36.6
	Obese (≥ 30)	292	33.4	38.0	28.7
Oil spill response worka	≥ 1 day	742	84.9	83.5	86.3
	None	132	15.1	16.5	13.7

^aParticipants reported working on Deepwater Horizon oil spill cleanup for at least one day in 2010–2011.

Based on the distribution of ambient styrene concentrations among Gulf region participants (Figure 1), we modeled neurologic symptoms in relation to quartiles of ambient estimated styrene. The distribution features a prominent right skew, with the top quartile ranging from 0.03 to 1.70 µg/m³ styrene. The referent exposure group (first quartile) experienced estimated ambient styrene levels up to 0.01 µg/m³ styrene. When comparing airborne styrene concentrations in the Gulf region to the entire U.S., the distributions are very similar with some slight variability in the top 5% of census tracts (data not shown). The distribution of blood styrene concentration has a less pronounced right skew (Figure 1) and the sample size is relatively small. We elected to dichotomize blood concentrations at the median value, 0.067 ng/mL styrene. We did not observe significant correlations between estimated ambient and measured blood styrene levels.

Figure 1.

Probability density of styrene concentrations in air (N=21,962) and blood (N=874).

Ambient styrene concentrations are National Air Toxics Assessment (NATA) 2011 modeled estimates of annual average concentrations (µg/m³) at the census tract level. Blood styrene exposure concentrations (ng/mL) are measured from a single blood draw obtained in the participant’s home.

Values at the top of reference lines indicate exposure concentrations; labels at the bottom of reference lines indicate locations in the exposure distribution: P25, 25^th percentile; P50, 50^th percentile; P75, 75^th percentile; P90, 90^th percentile; P95, 95^th percentile; Max, maximum value.

Increasing ambient styrene concentration was associated with the presence of any neurologic symptom, with statistically significant associations at each quartile (Table 2, Figure 2). For any CNS symptom, multiple CNS symptoms, any PNS symptom, and multiple PNS symptoms, the top quartile of estimated styrene was positively and significantly associated with reporting neurologic symptoms. The strongest association we observed among the cluster analyses was a 26% increase in prevalence of multiple CNS symptoms among those in the top quartile of ambient estimated styrene concentration (PR, 1.26; 95% CI: 1.15, 1.38). We observed significant linear trends (p-value < 0.01) between styrene quartiles and each of the symptom clusters (Table 2). Upon further examination in stratified analyses, we observed the strongest associations among non-White participants, with African Americans and those identifying as ‘Other races’ driving most associations between estimated ambient styrene and neurologic symptom clusters (Figure 3). This heterogeneity persisted, and results were unchanged, in models adjusted for income among the subset of participants for whom income is known (n=19,941).

Table 2.

Associations between estimated ambient styrene concentration and individual neurologic symptoms (N=21,962).

Outcome	Prevalence, N (%)	Ambient styrene	PR (95% CI)	p for trend
Any neurologic symptom	6,759 (30.8)	Q2	1.08 (1.03,1.14)	0.00003
		Q3	1.07 (1.01,1.12)
		Q4	1.12 (1.07,1.18)
Any CNS symptoma	4,690 (21.4)	Q2	1.05 (0.99,1.12)	<0.00001
		Q3	1.11 (1.05,1.18)
		Q4	1.17 (1.11,1.25)
≥ 2 CNS symptoms	2,037 (9.3)	Q2	1.12 (1.01, 1.24)	<0.00001
		Q3	1.16 (1.05,1.28)
		Q4	1.26 (1.15,1.38)
Dizziness	1,468 (6.7)	Q2	1.03 (0.91,1.17)	<0.00001
		Q3	1.12 (1.00,1.26)
		Q4	1.33 (1.19,1.48)
Nausea	1,046 (4.8)	Q2	1.04 (0.90,1.20)	0.007
		Q3	0.96 (0.83,1.11)
		Q4	1.24 (1.09,1.41)
Headache	3,121 (14.2)	Q2	1.10 (1.02,1.20)	<0.00001
		Q3	1.18 (1.09,1.27)
		Q4	1.21 (1.12,1.30)
Any PNS symptomb	4,014 (18.3)	Q2	1.09 (1.01,1.17)	<0.00001
		Q3	1.06 (0.99,1.14)
		Q4	1.16 (1.09,1.25)
≥ 2 PNS symptoms	2,429 (11.1)	Q2	1.09 (0.99,1.20)	0.0002
		Q3	1.14 (1.04,1.24)
		Q4	1.19 (1.08,1.30)
Tingling/Numbness	3,146 (14.3)	Q2	1.09 (1.00,1.18)	0.0001
		Q3	1.10 (1.01,1.19)
		Q4	1.18 (1.09,1.28)
Stumbling	739 (3.4)	Q2	1.20 (1.01,1.44)	0.0002
		Q3	1.20 (1.01,1.43)
		Q4	1.39 (1.18,1.64)
Blurred vision	1,656 (7.5)	Q2	1.16 (1.03,1.31)	<0.00001
		Q3	1.18 (1.04,1.32)
		Q4	1.43 (1.27,1.59)
Fatigue	3,529 (16.1)	Q2	1.13 (1.04,1.22)	0.001
		Q3	1.07 (0.99,1.16)
		Q4	1.16 (1.08,1.25)
Insomniac	1,552 (13.8)	Q2	1.19 (1.05,1.35)	0.001
		Q3	1.20 (1.05,1.36)
		Q4	1.25 (1.10,1.42)
Vomitc	1,363 (9.3)	Q2	1.18 (1.03,1.35)	0.001
		Q3	1.21 (1.06,1.38)
		Q4	1.24 (1.09,1.41)
Seizurec	263 (1.8)	Q2	1.14 (0.82,1.59)	0.001
		Q3	1.20 (0.87,1.65)
		Q4	1.62 (1.20, 2.19)

Models adjusted for sex, age, season, race, education, employment status at enrollment, alcohol drinking status at enrollment, and smoking status at enrollment.

Prevalence indicates total prevalence in the study sample (all estimated styrene levels).

Ambient styrene quartiles are based on National Air Toxics Assessment (NATA) 2011 estimates of annual average concentrations at the census tract level; referent group is the lowest quartile (not shown in table). Q2, second quartile; Q3, third quartile; Q4 fourth (highest) quartile.

^aCNS symptoms: dizziness, headache, nausea, sweating, palpitations.

^bPNS symptoms: tingling/numbness, blurred vision, and stumbling.

^cSample sizes vary for insomnia (n=11,236), vomiting (n=14,591), and seizure (n=14,591) because they were added to the interview after data collection was underway, and they are therefore ineligible for inclusion in symptom clusters (Any neurologic/CNS/PNS).

Figure 2.

Associations between estimated ambient styrene concentration and neurologic symptom clusters (N=21,962).

Models adjusted for sex, age, season, race, education, employment status at enrollment, alcohol drinking status at enrollment, and smoking status at enrollment.

Estimated ambient styrene quartiles are based on National Air Toxics Assessment (NATA) 2011 estimates of annual average concentrations at the census tract level; referent group is the lowest quartile (not shown on figure). Q2, second quartile; Q3, third quartile; Q4 fourth (highest) quartile. CNS symptoms include: dizziness, headache, nausea, sweating, and palpitations.

PNS symptoms include: tingling/numbness, blurred vision, and stumbling.

Figure 3.

Associations between estimated ambient styrene concentration and neurologic symptom clusters, stratified by race (N=21,962).

Models adjusted for sex, age, season, education, employment status at enrollment, alcohol drinking status at enrollment, and smoking status at enrollment.

Estimated ambient styrene quartiles are based on National Air Toxics Assessment (NATA) 2011 estimates of annual average concentrations at the census tract level; referent group is the lowest quartile (not shown on figure). Q2, second quartile; Q3, third quartile; Q4 fourth (highest) quartile. CNS symptoms include: dizziness, headache, nausea, sweating, and palpitations.

PNS symptoms include: tingling/numbness, blurred vision, and stumbling.

The prevalence of individual neurologic symptoms ranged from 1.8% for seizures to 16.1% for fatigue, with 30.8% of participants reporting at least one of the original neurologic symptoms (excluding vomiting, seizure, and insomnia) (Table 2). Among those participants who were also asked about the additional symptoms added later (n=16,536), the prevalence of any original neurologic symptom was 33.0% (and 38.0% including vomiting, seizure, and insomnia). The highest quartile of ambient estimated styrene was positively and significantly associated with each individual neurologic symptom. Many of the associations with individual symptoms demonstrated clearly increasing exposure-response relationships across quartiles of estimated styrene, and linear trends were significant (p < 0.01) for all symptoms.

In analyses of blood styrene above the median concentration and neurologic symptom clusters, we observed small to modest, nonsignificant positive associations for any PNS and multiple PNS symptoms (Table 3). The strongest association was a 37% increase in prevalence of multiple PNS symptoms, which was of borderline significance (PR, 1.37; 95% CI: 0.99, 1.89). Among CBS participants, seizure was the least common symptom (2.8%) and fatigue was the most prevalent (24.7%). Although associations between blood styrene and symptom clusters were generally not apparent, we did observe a noteworthy, albeit nonsignificant, positive association for the individual symptoms of dizziness (PR, 1.23; 95% CI: 0.77, 1.96), and a significant association for nausea (PR, 1.78; 95% CI: 1.04, 3.03) (Table 3).

Table 3.

Associations between blood styrene concentration and individual neurologic symptoms (N=874).

Outcome	Prevalence, N (%)	PR (95% CI)
Any neurologic symptom	434 (49.7)	1.01 (0.88,1.16)
Any CNS symptoma	274 (31.4)	0.97 (0.81,1.17)
≥ 2 CNS symptoms	128 (14.7)	1.10 (0.79,1.54)
Dizziness	90 (10.3)	1.23 (0.77,1.96)
Nausea	80 (9.2)	1.78 (1.04,3.03)
Headache	186 (21.3)	1.06 (0.82,1.37)
Any PNS symptomb	240 (27.5)	1.13 (0.91,1.42)
≥ 2 PNS symptoms	152 (17.4)	1.37 (0.99,1.89)
Tingling/numbness	190 (21.7)	1.07 (0.82,1.40)
Stumblingc	45 (5.2)	―
Blurred vision	105 (12.0)	0.97 (0.68,1.38)
Fatigue	216 (24.7)	0.96 (0.75,1.23)
Insomniad	130 (16.7)	0.97 (0.70,1.35)
Vomitingd	112 (13.1)	1.01 (0.72,1.44)
Seizurec,d	24 (2.8)	―

Models adjusted for sex, age, season, race, education, employment status at enrollment, alcohol drinking status at enrollment, smoking status at enrollment, and duration (days) between enrollment and blood draw.

Blood styrene exposure is classified as above the median concentration (0.067 ng/mL) compared to all measurements below the median concentration.

^aCNS symptoms include: dizziness, headache, nausea, sweating, and palpitations.

^bPNS symptoms include: tingling/numbness, blurred vision, and stumbling.

Prevalence indicates total prevalence in the study sample (both exposed and unexposed).

^cNo effect estimate presented for stumbling, seizure due to lack of model convergence.

^dSample sizes vary for insomnia (n=780), vomiting (n=857), and seizure (n=857) because they were added to the interview after data collection was underway, and they are therefore ineligible for inclusion in symptom clusters (Any neurologic/CNS/PNS).

Our findings were robust to exclusions and adjustments made in sensitivity analyses. We did not detect changes in results due to spatial clustering when we modeled associations specifying a random effect for census tract in models. These models, however, demonstrated some minor losses in precision. When restricting to never smokers, we observed stronger effects compared to those detected in the overall population. Accounting for oil spill cleanup work did not impact results, with interpretations unchanged whether models were adjusted for or stratified by cleanup participation. The interaction between styrene levels and cleanup work was consistently statistically non-significant. Likewise, the interaction between styrene levels and sex was non-significant. Point estimates were similar between men and women, although trends were more apparent in men due to improved precision from a larger sample size. Additional adjustments made to capture variability in socioeconomic status, such as income and concerns over affording housing and food, did not change results. Small sample sizes precluded our ability to evaluate heterogeneity by race in blood styrene analyses. Subgroup analyses were underpowered and unstable, with incomplete model convergence.

Spearman correlation coefficients between estimated ambient concentrations of styrene and each of benzene, toluene, and PM 2.5 were 0.69, 0.76, and 0.43, respectively. Associations between styrene and neurologic symptom clusters were stronger in models adjusted for estimated ambient benzene or toluene (Supplemental Figure 1). Adjusting for PM 2.5 had little effect, yielding estimates that were virtually identical to our primary results. Adjustment for blood levels of benzene and toluene did not meaningfully impact results for blood styrene exposure analyses (data not shown).

To evaluate the effect of depression on symptom reporting, we estimated associations separately for participants who reported a diagnosis of depression (n=3,069, 14%) and those who did not (n=18,805, 86%). Associations among participants without depression, representing the majority of the sample, were similar to overall results (Supplemental Figure 2). Similarly, adjustment for a range of chronic conditions (diabetes, cancer, asthma, emphysema, and obesity) did not elicit changes in overall results. Neither excluding participants who reported a total number of symptoms exceeding a range of thresholds (e.g. the 95^th percentile, or 12 symptoms), nor adjusting for total number of symptoms reported, had an appreciable impact on observed associations between ambient styrene and neurologic outcomes.

Restricting the duration of time between enrollment and blood draw to below the median (100 days) attenuated some associations between blood styrene and CNS symptoms slightly but did not meaningfully change associations. Neurologic effects of blood styrene were unchanged between models that were adjusted and unadjusted for the duration of time between enrollment and blood draw. Results were generally similar in analyses using other exposure contrasts for both estimated airborne and measured blood styrene levels (e.g. tertiles, quintiles, 90^th percentile, etc.), although risk estimates became more unstable as exposure groups became smaller. Excluding sweating and palpitations from the CNS symptom cluster did not meaningfully change results, though most associations were slightly stronger using this reduced CNS symptom cluster.

When we repeated analyses for clusters, as well as individual symptoms, among the subgroup of participants who completed the interview after questions about seizures, insomnia, and vomiting were added, results were virtually identical to originally reported associations. Indeed, because the order in which participants were enrolled in the study was largely random, we have no reason to suspect that individuals who responded before the additional symptoms were included would be meaningfully different from those who responded after their inclusion.

DISCUSSION

To our knowledge, this study was the first to assess subclinical neurotoxicity of styrene at environmental levels relevant to the general population. We observed consistent, positive relationships between increasing ambient estimated styrene levels and neurologic symptoms. Almost all associations were statistically significant, and a monotonic exposure-response was evident for each outcome. Although associations between blood styrene and neurologic symptom clusters were generally not apparent, we did observe suggestive effects for the PNS cluster and some individual symptoms (dizziness and nausea).

Among the analyses of estimated ambient styrene, we observed consistent associations with an unambiguous exposure-response and significant linear trends for the CNS cluster. These findings are supported by ATSDR’s report identifying the CNS as the primary target of styrene toxicity ¹, as well as increased mortality from diseases of the CNS, especially epilepsy, associated with styrene exposure in a cohort study of reinforced-plastics industry workers ⁵³.

In race-stratified analyses of ambient styrene and symptom clusters, associations were attenuated in White participants and strengthened among participants identifying as African American and Other races. We suspect that this heterogeneity is due, at least in part, to unmeasured confounding by socioeconomic status. This disparity may operate through cultural differences in symptom reporting, differential access to healthcare, unmeasured styrene exposure opportunities, and other residual socioeconomic inequalities. Consistent with previous research demonstrating that census tracts with higher proportions of racial minority and low-income populations tend to have higher NATA concentrations of air toxics ^54,55, African Americans in our study were most likely to live in high styrene census tracts and report lower incomes. We hypothesize that non-White participants in the present study may experience increased styrene exposure due to lower indoor air quality and increased proximity to intermittent source emissions that are not captured by NATA estimates. Further, there is evidence that synergistic health effects of social and physical environmental conditions, such as living in communities characterized by both elevated chemical exposures and social stressors, may result in increased health disparities ⁵⁶.

When examining individual neurologic symptoms, we observed the strongest associations for ambient styrene with reported seizures, blurred vision, stumbling, dizziness, and insomnia. Symptoms with the strongest evidence for styrene-induced neurotoxicity at occupational levels include blurred or distorted vision ^27,57–61, dizziness or lightheadedness ^10,28,38, headaches ^26,28,40, and fatigue ^26,28,38,40. Less consistent associations have been reported for vestibular impairment ⁴⁰, tingling and numbness ^8,28,30, seizures ^30,62, and insomnia ⁶³. Despite much lower ambient styrene levels in our study than those present in occupational settings, our results are generally consistent with these findings.

Blood styrene levels among CBS participants were two to three times higher than those measured in the National Health and Nutrition Examination Survey, for smokers and nonsmokers alike. Occupational styrene exposure levels are typically eight to twelve times higher than those measured in CBS. We observed suggestive associations of blood styrene concentration in relation to dizziness, nausea, and having multiple PNS symptoms. These effect estimates were large in magnitude, but some lacked precision due to the sample size and frequency of symptom reporting.

Having multiple PNS symptoms was associated with both ambient and blood concentrations of styrene. Dizziness and nausea also demonstrated some consistency between elevated ambient and blood styrene. Differences in associations by exposure measure may be attributable to the relevant temporal window each metric captures. Measured blood styrene generally reflects the previous 24 hours, whereas NATA estimated concentrations are annual averages thought to reflect longer term levels. Because blood samples were drawn, on average, 100 days after symptom information was collected, resulting exposure misclassification is especially likely to have limited our ability to detect significant associations with blood styrene.

Our study is the first to investigate whether neurologic effects observed among highly exposed occupational populations are also observed among individuals with lower levels of styrene exposure more typical of the general population. Recent environmental studies evaluating effects of simultaneous exposure to multiple hazardous air pollutants (HAPs) have documented associations between environmental styrene and clinical neurologic outcomes. In studies evaluating modeled ambient concentrations of a variety of HAPs, styrene was associated with increased risk of autism spectrum disorder ^64,65 and amyotrophic lateral sclerosis ⁶⁶. These results implicate low-level, chronic styrene exposure as a possible public health problem. However, a cross-sectional analysis of blood VOCs and neurobehavioral testing found a general lack of significant adverse effects, with the exception that a mixture of BTEX and styrene was modestly associated with slower reaction time ⁶⁷.

Studies in humans and experimental in vitro and in vivo animal models have attempted to determine the mode of action for styrene neurotoxicity, with a dopaminergic mechanism gaining traction ⁶⁸, but explanations remain speculative. Several studies suggest that styrene exposure alters dopamine metabolism, marked by decreased dopamine levels and increased dopamine receptors in rodents and humans ^69–71. The styrene metabolites phenylglyoxylic acid and mandelic acid are shown to deplete dopamine in neurologic tissues ⁷². This mechanism is corroborated in blood samples of styrene-exposed plastics workers for whom prolactin levels are elevated, as prolactin release is chronically inhibited by dopamine ⁷³. Consistent with disturbance of the dopaminergic functions of the brain, styrene exposure potentiates a dose-dependent decrease in brain dopamine in male rats ⁷¹. Styrene also causes cell loss and dopamine depletion in retinas isolated from female rats ⁷⁴, supporting the established association between occupational styrene exposure and impaired vision ⁷⁵.

Our study has several strengths, including a large sample size, a well-characterized, diverse, understudied population, multiple sensitive measures of neurotoxicity, complementary metrics of exposure, and findings that were robust to multiple sensitivity analyses.

Neurologic symptoms can reveal subtle impairments in neurologic function before the occurrence of clinically apparent disease. While less severe, these symptoms may be more sensitive to lower exposure levels, longer lasting, and more prevalent in the general population. Self-reported symptoms can provide highly sensitive measures of toxicant-associated neurotoxicity ⁷⁶. Thus, symptom assessment is an appropriate measure to capture the potentially subtle, widespread neurotoxic effects of the observed environmental styrene levels.

We assessed two metrics of styrene exposure, one reflecting typical ambient levels and one reflecting short-term internal burden. Modeling exposure this way helps address the trade-offs between possible confounding bias for blood styrene, and potential measurement error of NATA ambient styrene estimates ⁷⁷. Associations were not entirely consistent between measured blood and estimated ambient styrene, which may be due to limitations of both biomarkers and proxy exposure measures, or the temporal windows they reflect. NATA estimates have been successfully used as measures of human air pollution exposure in epidemiologic studies of cancer ^78–80, asthma ⁸¹, birth defects ⁸², autism spectrum disorder ^64,65,83, and neurodegenerative diseases ⁶⁶. We used NATA styrene estimates as indicators of typical environmental exposure, and blood styrene measurements to capture internal burden resulting from recent exposures. Blood styrene is a validated biomarker specific to styrene exposure ⁸⁴ and it has been used extensively in occupational research ³, as well as in general population monitoring ¹.

Limitations of our study include the use of an estimated annual average ambient concentration, single blood draw for styrene measurement, and lack of detailed medical treatment and occupational exposure information. Due to the cross-sectional study design, we cannot confirm temporality between styrene exposure and the onset of neurologic symptoms. It is, however, unlikely that the symptoms assessed would lead to the estimated ambient or blood styrene concentrations we observed. More relevant to the cross-sectional design limitation is our inability to distinguish between transient effects due to acute exposures and persistent neurologic effects of long-term styrene exposure.

Routine monitoring of ambient styrene in the U.S. does not provide sufficient temporal or spatial coverage to support exposure interpolation methods ^85–87. NATA remains the only spatially-referenced exposure data source with sufficient geographic coverage for the Gulf state region. We recognize that using this type of surrogate for typical ambient exposure does not incorporate variability in residential history (e.g., duration at the assigned census tract), daily activity patterns, time spent in/outdoors, or time spent outside of the residential census tract. Our inability to account for these factors and other assumptions inherent to an annual average estimate of air pollution limits interpretation for acute exposure scenarios, but NATA data are considered a valid estimation of usual exposure levels. If styrene-associated neurologic effects are the result of long-term exposure, this type of exposure metric may be better suited to capturing the associations of interest. Long-term ambient styrene trends indicate that year-to-year regional variation in concentration is not substantial, suggesting that the annual average estimate is an appropriate measure of usual, long-term exposure ¹⁸. Stable blood styrene levels in independent U.S. population samples, measured cross-sectionally, over a 20-year period corroborate these findings ^16,84,88.

Although we obtained only a single blood measurement from each individual, the sample size is large for a biomarker-based study. At occupational exposure levels, biomarkers consistently predict exposure classification according to job title and occupational styrene-related activities ^{4,7,26,89–97}. In general, these biomarkers correlate well with indoor air styrene levels, but the relationship does not necessarily persist for outdoor air styrene concentrations. Indeed, we observed limited concordance between measured blood styrene and NATA estimated concentrations (Werder EJ, submitted). In occupational studies, the correlation between styrene in blood and indoor air ranges from 0.62 to 0.94 ^11,96,98. In a study of low occupational styrene levels measured in air, blood, and urine, the highest correlations with ambient levels were observed for blood styrene ⁹⁵.

We lacked the data necessary to account for participants’ use of medical treatments and medication. As such, neurologic symptoms may have been masked or attenuated by medical treatment, thus increasing outcome measurement error. We did, however attempt to account for general health status by evaluating depression, chronic conditions, body mass index, diabetes, and number of symptoms reported.

Lacking detailed information on styrene-specific occupational exposure opportunities, we assessed reported industry, occupation, and activity information for participant’s recent and longest-held jobs. Ultimately, occupational styrene exposure opportunities were so rare in our study population (0.1%), that we concluded that occupational exposures were not appreciably influencing results. Although oil spill cleanup is not considered a source of occupational styrene exposure, we did consider the potential fundamental differences between participants who participated in oil spill cleanup and those who did not. These analyses did not reveal any impacts of cleanup worker status on the association between styrene and neurologic symptoms.

Our results provide strong evidence for an association between ambient styrene exposure and neurologic symptoms, suggesting that styrene may be neurotoxic at environmental levels relevant to the general population. These findings were consistent across a wide range of sensitivity analyses. The relationship between blood styrene exposure and neurologic symptoms was equivocal, but suggestive of an association for certain endpoints. Although we cannot refute the possibility of acute effects, these results are more indicative of a relationship between typical cumulative styrene exposure and neurologic effects. Due to the cross-sectional nature of our study, the temporality of exposure and outcome is uncertain. While short-term acute intoxication is typically reversible, ceasing when styrene is cleared from the body, more concerning for overall health are the potential chronic, subtle but demonstrable, and irreversible effects that persist after styrene is cleared from the body ⁹⁹. Timing of environmental styrene exposure and duration of neurotoxic effects may be an important area of future public health research.

HIGHLIGHTS.

• Increasing ambient styrene is consistently associated with neurologic symptoms

• The strongest associations are among non-White participants

• Associations between blood styrene and neurologic symptoms are only suggestive