Exposure to 1,3-Butadiene in the U.S. Population: National Health and Nutrition Examination Survey 2011–2016

Abstract

1,3-Butadiene is a volatile organic compound with a gasoline-like odour that is primarily used as a monomer in the production of synthetic rubber. The International Agency for Research on Cancer has classified 1,3-butadiene as a human carcinogen. We assessed 1,3-butadiene exposure in the U.S. population by measuring its urinary metabolites N-acetyl-S-(3,4-dihydroxybutyl)-L-cysteine (34HBMA), N-acetyl-S-(1-hydroxymethyl-2-propenyl)-L-cysteine (1HMPeMA), N-acetyl-S-(2-hydroxy-3-butenyl)-L-cysteine (2HBeMA), and N-acetyl-S-(4-hydroxy-2-buten-1-yl)-L-cysteine (4HBeMA). Urine samples from the 2011 to 2016 National Health and Nutrition Examination Survey were analysed for 1,3-butadiene metabolites using ultrahigh-performance liquid chromatography/tandem mass spectrometry. 34HBMA and 4HBeMA were detected in >96% of the samples; 1HMPeMA and 2HBeMA were detected in 0.66% and 9.84% of the samples, respectively. We used sample-weighted linear regression models to examine the influence of smoking status (using a combination of self-reporting and serum-cotinine data), demographic variables, and diet on biomarker levels. The median 4HBeMA among exclusive smokers (31.5 μg/g creatinine) was higher than in non-users (4.11 μg/g creatinine). Similarly, the median 34HBMA among exclusive smokers (391 μg/g creatinine) was higher than in non-users (296 μg/g creatinine). Furthermore, smoking 1–10, 11–20, and >20 cigarettes per day (CPD) was associated with 475%, 849%, and 1143% higher 4HBeMA (p < 0.0001), respectively. Additionally, smoking 1–10, 11–20, and >20 CPD was associated with 33%, 44%, and 102% higher 34HBMA (p < 0.0001). These results provide significant baseline data for 1,3-butadiene exposure in the U.S. population, and demonstrate that tobacco smoke is a major exposure source.

1. Introduction

1,3-Butadiene is a volatile organic compound (VOC) with a gasoline-like odour that is primarily used as a monomer in the production of synthetic rubber (ATSDR 2012, OSHA 2012). This compound is also used in the polymer production of styrene-butadiene rubber and acrylonitrile-butadiene-styrene resin plastics (EPA 2002, ATSDR 2012). Environmental sources of 1,3-butadiene include industrial emissions; automobile exhaust; burning of wood, plastics, and rubber; cigarette smoke (ATSDR 2012); and cooking emissions (Huang et al. 2020). Additionally, 1,3-butadiene has been found in plastic containers and selected foods (Yurawecz et al. 1976, McNeal and Breder 1987, Abrantes et al. 2000). Inhalation is the main route of exposure. Exposure by ingestion is unlikely since 1,3-butadiene rapidly evaporates to the atmosphere. 1,3-Butadiene is poorly soluble in water; however, low levels are released to water and soil (EPA 2002, ATSDR 2012). In the United States, the reference concentration for breathing 1,3-butadiene in air is 0.9 ppb (EPA 2002); the legal occupational exposure limit is set to 1 ppm for an 8-h workday and a short-term exposure limit of 5 ppm for 5 minutes (OSHA 1996). Ambient 1,3-butadiene undergoes photo-initiated chemical breakdown and is expected to have a half-life of approximately 6h (ATSDR 2012), or 2–10 h as estimated from inhalation studies in animals (Bond et al. 1987). Nonetheless, low levels of 1,3-butadiene are constantly present in urban and suburban areas due to automobile exhaust, biomass burning, or industrial emissions (Hendler et al. 2010, ATSDR 2012, Gallego et al. 2018, Xiong et al. 2020).

Human studies indicate that inhalational absorption of 2 ppm 1,3-butadiene for 20 minutes varied from 18% to 74% (Lin et al. 2001). In the liver, cytochrome P450 oxidises 1,3-butadiene to highly reactive butadiene monoepoxide (BMO). Enzymatic hydrolysis of BMO by epoxide hydrolase enzyme leads to 1,2-dihydroxy-3-butene (butadiene-diol). Conjugation with glutathione produces mercapturic acids that are excreted in urine (Csanady et al. 1992, Richardson et al. 1998, 1999). Urinary biomarkers used to evaluate 1,3-butadiene exposure include N-acetyl-S-(3,4-dihydroxybutyl)-L-cysteine (34HBMA) and the isomeric mixture of N-acetyl-S-(1-hydroxymethyl-2-propenyl)-L-cysteine (1HMPeMA), N-acetyl-S-(2-hydroxy-3-butenyl)-L-cysteine (2HBeMA), and N-acetyl-S-(4-hydroxy-2-buten-1-yl)-L-cysteine (4HBeMA). The chemical structures of 1,3-butadiene and its urinary metabolites are depicted in Scheme 1. Hydrolysis of BMO leads to the formation of butadiene-diol which subsequently reacts with glutathione to form 34HBMA, while 1HMPeMA, 2HBeMA and 4HBeMA are formed by direct conjugation of BMO with glutathione (Csanady et al. 1992, Bechtold et al. 1994, Richardson et al. 1998, 1999, Urban et al. 2003).

Scheme 1.

Chemical structures of 1,3-butadiene and its urinary metabolites.

1,3-Butadiene is classified as a human carcinogen by the International Agency for Research on Cancer, based on animal studies as well as epidemiological studies of humans with occupational and environmental exposures (IARC 2008). Importantly, 1,3-butadiene is a harmful constituent in tobacco smoke (Brunnemann et al. 1989, Smith et al. 2000, Fowles and Dybing 2003, IARC 2004, FDA 2012, Haussmann 2012, Soeteman-Hernandez et al. 2013). It is associated with cancer susceptibility in humans and animals (Boldry et al. 2017, Murphy et al. 2018, Etemadi et al. 2020). In laboratories, rodents exposed to 1,3-butadiene developed tumours in the lungs, including alveolar/bronchiolar adenomas (NTP 1984, Huff et al. 1985, Melnick et al. 1990, NTP 1993). Tobacco use is considered a global epidemic and is the leading cause of preventable morbidity and mortality in the United States (CDC 2014, HHS 2014, WHO 2017, Creamer et al. 2019). According to the U.S. Department of Health and Human Services, cigarette smoking was responsible for more than 480,000 premature deaths annually from 2005 to 2009 (HHS 2014). Additionally, reports indicate that constituents in smoke, such as 1,3-butadiene, may contribute to health risks associated with exposure to second-hand smoke (SHS), which causes adverse health effects in infants, children, and nonsmokers (HHS 2014, St Helen et al. 2014).

To date, no studies have been published examining human exposure to 1,3-butadiene on a population-wide scale, despite its harmful health impact. We examined urinary biomarkers of 1,3-butadiene exposure in participants of the 2011–2016 National Health and Nutrition Examination Survey (NHANES) to obtain population-based biomonitoring data of the U.S. civilian, non-institutionalized population. We used multiple linear regression models to evaluate the influence of smoking status, demographic variables, and diet on 1,3-butadiene exposure. Here, we present analysis of urinary 4HBeMA and 34HBMA to assess exposure to 1,3-butadiene in the U.S. general population.

2. Materials and methods

2.1. Study design

NHANES is a population-based survey designed to assesses the health and nutritional status of the civilian, non-institutionalized U.S. population based on data collected from questionnaires, physical examinations, and biological samples. This cross-sectional study is conducted by the National Centre for Health Statistics (NCHS) of the U.S. Centres for Disease Control and Prevention (CDC) (CDC 2011–2017). The study protocol was reviewed and approved by the NCHS Research Ethics Review Board (ERB) (NCHS ERB Approval Protocols 2011–2017). Informed written consents were obtained from all subjects before they participated in the survey. A total of 29,902 participants aged 3 years and older provided spot urine samples for the 2011–2016 NHANES, and we quantified 4HBeMA, 34HBMA, 1HMPeMA and 2HBeMA in a one-third subset.

Laboratory data for 7677 participants were reported for 4HBeMA and 34HBMA (NHANES datasets: UVOC). Participants were excluded from analysis if they did not meet the criteria for either exclusive smoker or non-user (N= 980), for missing serum cotinine data (N=214), for missing creatinine data (N=4) or for missing data for other variables used in the regression model (N=582). This assignment left 5897 participants eligible for statistical analysis.

In this report, study participants were identified as exclusive smokers (N=726) or exclusive daily users of cigarette products, if they responded “yes” to NHANES question SMDANY (tobacco use within five days prior to NHANES physical examination), “yes” to SMQ690A (cigarette use), and “no” to SMQ690B-SMQ690J (use of pipes, cigars, chewing tobacco, snuff, patch/gum, hookah/water pipes, e-cigarettes, snus, and dissolvable tobacco), according to NHANES questionnaire data on recent tobacco use (NHANES datasets: SMQRTU). The use of tobacco products among exclusive smokers was confirmed by a serum cotinine >10mg/mL (NHANES datasets: COT). Subsequently, participants were identified as non-users (N=5171) if they answered “no” to SMDANY or had serum cotinine ≤10mg/mL. A serum-cotinine threshold of >10mg/mL has been identified as consistent with active use of combusted cigarette product (Pirkle 1996) and was used to stratify self-identified exclusive smokers and non-users in our statistical analyses.

2.2. Laboratory analysis

Spot urine samples were assayed for urinary 4HBeMA, 34HBMA, 1HMPeMA, and 2HBeMA by running a 50 μL aliquot of each sample through an ultra-high-performance liquid chromatograph (UPLC; I-Classic Acquity, Waters Inc., Milford, MA) coupled with electrospray ionisation (ESI) tandem mass spectrometry (ESI-MS/MS; Sciex 5500 Triple quad, Sciex, Framingham, MA) as described previously (Alwis et al. 2012). A detailed description of the materials and the analytical method used can be found elsewhere (Alwis et al. 2012). The acronyms used in this paper for the biomarkers of exposure to 1,3-butadiene differ from Alwis et al. and are based on the newly established guidelines on VOCM acronym harmonisation (Tevis et al., submitted). Sample concentrations were determined based on their relative response ratio (ratio of native analyte to stable isotope-labelled internal standard) against a calibration curve with known standard concentrations. For 34HBMA, m/z 2 5 0 → 121 and m/z 2 5 0 → 75 were monitored as the quantitation and confirmation ion transitions, respectively. For monohydroxylated isomers, m/z 2 3 2 → 103 and m/z 2 3 3 → 103 were monitored as the quantitation and confirmation ion transitions, respectively. For labelled internal standards, we monitored ion transitions m/z 2 5 4 → 125 for 34HBMA-¹³C₄, m/z 2 3 8 → 109 for 1HMPeMA-d₆, m/z 2 3 6 → 103 for 2HBeMA-¹³C₃-¹⁵N, and m/z 2 3 5 → 103 for 4HBeMA-d₃. The limits of detection (LOD) were 5.25 mg/mL 34HBMA, and 0.7 mg/mL for 1HMPeMA and 2HBeMA, and 0.6 mg/mL 4HBeMA, respectively.

2.3. Statistical analysis

NHANES is a population-based survey developed to recruit participants through a multistage probability sampling design. In order to produce unbiased, nationally representative statistics with appropriate variance estimates, it is necessary to respect the design, applying sampling probability, and using estimation procedure involving Taylor series linearisation accounting for clustering. This necessity was addressed through the use of SURVEYREG and SURVEYMEANS subroutines of SAS® 9.4 statistical software (SAS® Institute). Sample-weighted multiple linear regression models (N=5897) stratified by cigarette use status (exclusive smokers vs. non-users) were fit to 2011–2016 NHANES data, where the dependent variables were urinary concentration of 34HBMA and 4HBeMA (mg/mL). The distribution of urinary biomarker measurements was strongly right-skewed, which could have adversely affected hypothesis testing, so we used natural log-transformed biomarker data for regression analysis. We report coefficients from these models along with their 95% confidence intervals (95% CI) and p-values. The statistical significance was set to α = 0.05. Additionally, to facilitate predictor interpretation, the exponentiated coefficient in the computer regression equation represents the proportional change of biomarker concentration as previously described (Biren et al. 2020). An evaluation of statistical reliability was performed to ensure all proportions followed the NCHS Data Presentation Standard.

Sample-weighted regression models were stratified by cigarette use, and the following self-reported variables were included as predictors: urinary creatinine (g/L), diet, sex, age, and race/Hispanic origin. Creatinine, a waste product of creatine and creatine phosphate (produced from muscle metabolism), is excreted in urine at a relatively constant rate (Barr et al. 2005). Age was categorised into the following ranges and is consistent with previous studies: 3–5, 6–11, 12–19, 20–39, 40–59, and ≥60 years (CDC 2011–2017, Bagchi et al. 2018, Capella et al. 2019, Espenship et al. 2019). An additional predictor, weight status [body mass index (BMI)], was calculated from measurements taken at the NHANES physical examination. Standard definitions for underweight (BMI < 18.5 kg/m²), healthy weight (18.5 ≤ BMI < 25), and overweight/obesity (BMI ≥ 25) apply to adults ≥ 20 years. Participants younger than 20 years were classified based on their BMI percentile for their sex and age: below the 5th percentile (underweight), between the 5th and 85th percentile (healthy weight), and above the 85th percentile (overweight/obesity).

The NHANES sample person questionnaire also collected information on the participant’s diet during the Mobile Examination Centre interviews. The participant’s recollection of the type and amount of food consumed was then standardised to a numerical coding system and food mass. Dietary exposure was investigated by assessing the mass participants consumed within each U.S. Department of Agriculture (USDA) food group for the 24-h period (midnight to midnight) preceding the day of the in-person dietary recall interview and urine collection. Data for the 24-h recall period are contained in the publicly available NHANES Individual Foods – First Day file (NHANES datasets: DR1IFF). This file lists each food, water, or beverage consumed by the participant, including the mass reported consumed and eight-digit USDA food code. Standardised hierarchical food groups can be identified from the USDA code, where the first digit represents one of nine major food groups, and each subsequent digit represents subgroups of increasing specificity (Ahuja et al. 2013). The mass consumed in each food group was summed so that each participant was represented by a single record describing their dietary intake for the previous 24 h. Each participant’s dietary intake was first apportioned over nine food groups: milk products; meat and poultry; eggs; legumes, nuts, and seeds; grain products; fruits; vegetables; fats, oils, and salad dressings; and sugars, sweets, and beverages.

Cotinine is a highly specific metabolite of nicotine, the primary addictive chemical in tobacco, and is used as a biomarker of environmental tobacco smoke exposure (Watts et al. 1990, Benowitz 1996). In this report, serum cotinine was used as a continuous predictor to evaluate tobacco smoke exposure for both exclusive smokers (N=726) and non-users (N=5171). Among non-users, tobacco smoke exposure is attributable to inhalation of SHS, which can be quantified with serum cotinine. In order to assess the association between urinary 1,3-butadiene metabolites and frequency of cigarette smoking, we performed an unstratified, sample-weighted regression model (N=5897) in which exposure among exclusive smokers was represented by the self-reported average number of cigarettes smoked per day (CPD) over the five days preceding the NHANES physical exam. This CPD regression model comprised the same predictors as the stratified models, except that tobacco smoke exposure was classified in the following mutually exclusive categories: ≤0.015 mg/mL serum cotinine and 0 CPD (unexposed to tobacco smoke), >0.015 – ≤10 mg/mL serum cotinine and 0 CPD (presumptively exposed to SHS), >10 mg/mL serum cotinine and 1–10 CPD, >10 mg/mL serum cotinine and 11–20 CPD, and >10 mg/mL serum cotinine and >20 CPD. The reference category was unexposed participants and was defined at ≤0.015 mg/mL serum cotinine. The analytic dataset for the CPD model and the stratified models comprised the same participants (N=5897).

3. Results

Urinary 4HBeMA and 34HBMA were detected in 99.9% and 96.7% of samples analysed, respectively. A Pearson correlation coefficient of 0.50 was observed between urinary 4HBeMA and 34HBMA (Supplemental Figure 1). The detection rate of 1HMPeMA (0.7%) and 2HBeMA (9.8%) was not sufficient for robust statistical analysis, thus those results were not further discussed in this report. Table 1 shows the sample-weighted distributions for demographic variables stratified by smoking status for the 5897 participants included in this report. We also present sample-weighted summary statistics categorised by sex, age, race/Hispanic origin, and weight status, among exclusive smokers and non-users for urinary 4HBeMA (Table 2) and 34HBMA (Table 3). Sample-weighted median 4HBeMA among exclusive smokers (31.5 μg/g creatinine) was considerably higher than for non-users (4.11 μg/g creatinine). Similarly, sample-weighted median 34HBMA among exclusive smokers (391 μg/g creatinine) was higher than for non-users (296 μg/g creatinine). Sample-weighted medians and selected percentiles are shown in Supplemental Table 1 for the interpretation of the percentage of 4HBeMA and 34HBMA changes for all numeric predictors.

Table 1.

Sample-weighted demographic distribution of 2011–2016 NHANES among exclusive smokers and non-users (N=5897)^a.

Predictor	Level	Nb, Exclusive smokers	Percent (SE)c, exclusive smokersd	Nb, non-users	Percent (SE)c, non-userse
NHANES Cycle	2011 – 2012	259	37.89 (2.42)	1698	34.79 (2.07)
NHANES Cycle	2013 – 2014	238	30.61 (2.19)	1669	32.68 (2.03)
NHANES Cycle	2015 – 2016	229	31.50 (2.49)	1804	32.54 (2.26)
Â	All	726	100 (0.00)	5171	100 (0.00)
Sex	Male	420	50.88 (2.87)	2467	46.99 (0.83)
Sex	Female	306	49.12 (2.87)	2704	53.01 (0.83)
Age	3–5	0	N/Af	239	0.83 (0.07)
Age	6–11	0	N/Af	776	8.46 (0.42)
Age	12–19	41	4.14 (0.63)	900	12.95 (0.70)
Age	20–39	269	38.28 (2.15)	1121	27.09 (1.09)
Age	40–59	267	42.87 (2.52)	1029	27.52 (1.05)
Age	≥60	149	14.71 (1.69)	1106	23.16 (1.05)
Race/Hispanic Origin	Non-Hispanic White	335	67.73 (3.20)	1666	62.85 (2.43)
Race/Hispanic Origin	Non-Hispanic Black	197	14.51 (1.87)	1119	10.61 (1.20)
Race/Hispanic Origin	Hispanic	132	12.04 (1.89)	1565	18.14 (1.86)
Race/Hispanic Origin	Other Race/Multi-Racial	62	5.72 (0.91)	821	8.40 (0.67)
Weight Status	Healthy Weight	244	31.92 (2.12)	2012	35.16 (1.29)
Weight Status	Overweight/Obesity	462	65.32 (2.02)	3066	63.41 (1.36)
Weight Status	Underweight	20	2.76* (0.79)	93	1.43 (0.21)

^aSame data as in stratified serum cotinine regression model.

^bSample size, not sample-weighted.

^cStandard error, sample-weighted.

^dSample-weighted participants reporting using cigarettes (and no other tobacco products) 5 days prior to physical examination and with serum cotinine measurement >10 mg/ml.

^eSample-weighted participants reporting not using cigarettes during 5 days prior to physical examination or with serum-cotinine measurement ≤10 mg/mL.

^fN/A: not applicable.

Table 2.

Sample-weighted median urinary 4HBeMA/MHBMA3 concentrations (μg/g creatinine) among 2011–2016 NHANES exclusive smokers and non-users (N=5897)^a.

Predictor	Level	Exclusive smokersb Median [25th, 75th percentile]	Non-usersc Median [25th, 75th percentile]
NHANES Cycle	2011–2012	37.0 [17.6, 62.1]	4.46 [2.85, 6.90]
NHANES Cycle	2013–2014	31.0 [15.6, 50.6]	4.62 [3.09, 7.20]
NHANES Cycle	2015–2016	28.9 [14.4, 47.9]	3.39 [2.44, 5.46]
Â	All	31.5 [16.6, 51.9]	4.11 [2.72, 6.55]
Sex	Male	27.0 [14.7, 43.0]	3.96 [2.67, 6.41]
Sex	Female	40.7 [18.9, 62.7]	4.26 [2.80, 6.70]
Age	3–5	N/Ad	7.04 [4.69, 10.8]
Age	6–11	N/Ad	6.14 [3.91, 8.69]
Age	12–19	17.0 [7.00, 36.6]	3.57 [2.39, 5.46]
Age	20–39	24.5 [13.3, 41.6]	3.62 [2.43, 5.90]
Age	40–59	38.7 [22.2, 63.7]	3.99 [2.70, 6.44]
Age	≥60	40.7 [20.4, 64.2]	4.62 [3.15, 6.66]
Race/Hispanic Origin	Non-Hispanic White	37.8 [20.3, 57.3]	4.24 [2.80, 6.61]
Race/Hispanic Origin	Non-Hispanic Black	21.5 [12.4, 34.0]	3.63 [2.54, 5.69]
Race/Hispanic Origin	Hispanic	20.3 [10.0, 38.5]	3.99 [2.65, 6.57]
Race/Hispanic Origin	Other Race/Multi-Racial	27.1 [15.2, 48.8]	3.99 [2.56, 6.62]
Weight Status	Healthy Weight	33.8 [16.9, 52.5]	4.43 [2.78, 6.93]
Weight Status	Overweight/Obesity	30.8 [16.0, 51.5]	3.95 [2.69, 6.23]
Weight Status	Underweight	33.3 [16.8, 49.2]	4.87 [3.26, 7.10]

^aSame data as in stratified serum cotinine regression models.

^bSample-weighted participants reporting using cigarettes (and no other tobacco products) 5 days prior to physical examination and with serum-cotinine measurement >10 mg/mL.

^cSample-weighted participants reporting not using cigarettes during 5 days prior to physical examination or with serum cotinine measurement ≤10 mg/mL.

^dN/A: not applicable.

Table 3.

Sample-weighted median urinary 34HBMA/DHBMA concentrations (μg/g creatinine) among 2011–2016 NHANES exclusive smokers and non-users (N=5897)^a.

Predictor	Level	Exclusive smokersb Median [25th, 75th percentile]	Non-usersc Median [25th, 75th percentile]
NHANES Cycle	2011–2012	382 [302, 489]	278 [220, 363]
NHANES Cycle	2013–2014	370 [281, 465]	277 [211, 364]
NHANES Cycle	2015–2016	411 [333, 546]	334 [263, 428]
Â	All	391 [304, 506]	296 [230, 385]
Sex	Male	370 [286, 472]	280 [220, 365]
Sex	Female	413 [331, 556]	309 [238, 404]
Age	3–5	N/Ad	576 [463, 786]
Age	6–11	N/Ad	390 [306, 493]
Age	12–19	319 [229, 369]	266 [218, 338]
Age	20–39	345 [282, 441]	256 [209, 322]
Age	40–59	409 [332, 570]	289 [227, 351]
Age	≥60	461 [371, 526]	357 [271, 446]
Race/Hispanic Origin	Non-Hispanic White	408 [330, 519]	310 [238, 406]
Race/Hispanic Origin	Non-Hispanic Black	319 [248, 428]	249 [199, 321]
Race/Hispanic Origin	Hispanic	369 [280, 474]	287 [226, 373]
Race/Hispanic Origin	Other Race/Multi-Racial	383 [299, 550]	279 [225, 357]
Weight Status	Healthy Weight	401 [312, 540]	310 [238, 411]
Weight Status	Overweight/Obesity	383 [302, 484]	290 [225, 373]
Weight Status	Underweight	404 [314, 534]	333 [247, 495]

^aSame data as in stratified serum cotinine regression models.

^bSample-weighted participants reporting using cigarettes (and no other tobacco products) 5 days prior to physical examination and with serum cotinine measurement >10 mg/mL.

^cSample-weighted participants reporting not using cigarettes during five days prior to physical examination or with serum cotinine measurement ≤10 mg/mL.

^dN/A: not applicable.

We studied the association of urinary 4HBeMA with tobacco use, diet, and demographic variables using three multivariate models: exclusive smokers (Table 4), non-users (Table 5), and all study participants (Table 6). We used stratified analysis to account for the potential confounding effect between smoking status and other covariates in the model. Tobacco smoke exposure was consistently associated with higher urinary 4HBeMA compared with non-users in all models, and this effect was dose dependent (Figure 1). Compared with non-users, smoking 1–10, 11–20, and >20 CPD was significantly associated with 475%, 849% and 1,143% higher 4HBeMA levels, respectively, controlling for other variables (all p < 0.0001, Table 6). We found a similar dose-dependent interaction when we modelled tobacco-smoke exposure using serum cotinine for both exclusive smokers (Table 4) and non-users (Table 5): higher serum cotinine was associated with higher urinary 4HBeMA (p < 0.0001 for both smokers and non-smokers with SHS exposure), controlling for other variables.

Table 4.

Sample-weighted multiple linear regression model for urinary 4HBeMA/MHBMA3 concentrations (mg/mL) among 2011–2016 NHANES exclusive smokers^a (N=726).

Predictor	Level	Coefficient [95% CI]b	p-Valuec	Exponentiated slope [95% CI]d
Intercept	Intercept	1.6847 [1.3372, 2.0321]	<.0001	5.39 [3.81, 7.63]
Creatinine, Urine [g/L]	Slope	0.5831 [0.5007, 0.6656]	<.0001	1.79 [1.65, 1.95]
Cotinine, Serum [mg/mL]	Slope	0.0029 [0.0021, 0.0038]	<.0001	1.00 [1.00, 1.00]
Sex	Male	Ref.e	.	Ref.e
Sex	Female	0.3291 [0.1588, 0.4994]	0.0003	1.39 [1.17, 1.65]
Age	3–5	N/A	.	N/A
Age	6–11	N/A	.	N/A
Age	12–19	−0.0316 [−0.2878, 0.2246]	0.8051	0.969 [0.750, 1.25]
Age	20–39	Ref.e	.	Ref.e
Age	40–59	0.2316 [0.1100, 0.3531]	0.0004	1.26 [1.12, 1.42]
Age	≥60	0.3537 [0.1451, 0.5622]	0.0013	1.42 [1.16, 1.75]
Race/Hispanic Origin	Non-Hispanic White	Ref.e	.	Ref.e
Race/Hispanic Origin	Non-Hispanic Black	−0.3690 [−0.5338, −0.2043]	<.0001	0.691 [0.586, 0.815]
Race/Hispanic Origin	Hispanic	−0.1129 [−0.3109, 0.0851]	0.2573	0.893 [0.733, 1.09]
Race/Hispanic Origin	Other Race/Multi-Racial	−0.0091 [−0.3497, 0.3316]	0.9576	0.991 [0.705, 1.39]
Weight Status	Underweight	0.0991 [−0.2599, 0.4582]	0.5812	1.10 [0.771, 1.58]
Weight Status	Healthy Weight	Ref.e	.	Ref.e
Weight Status	Overweight/Obesity	0.0703 [−0.0675, 0.2081]	0.31	1.07 [0.935, 1.23]
Food Consumed [kg/day]	Milk Products	−0.0154 [−0.2540, 0.2233]	0.8974	0.985 [0.776, 1.25]
Food Consumed [kg/day]	Meat, Poultry, Fish	0.0035 [−0.3414, 0.3485]	0.9836	1.00 [0.711, 1.42]
Food Consumed [kg/day]	Eggs	0.3733 [−0.6180, 1.3646]	0.4525	1.45 [0.539, 3.91]
Food Consumed [kg/day]	Legumes, Nuts, Seeds	−0.1376 [−1.3804, 1.1053]	0.8248	0.871 [0.251, 3.02]
Food Consumed [kg/day]	Grain Products	−0.0547 [−0.3304, 0.2210]	0.6915	0.947 [0.719, 1.25]
Food Consumed [kg/day]	Fruits	−0.0775 [−0.3616, 0.2065]	0.5854	0.925 [0.697, 1.23]
Food Consumed [kg/day]	Vegetables	0.0849 [−0.3927, 0.5624]	0.7223	1.09 [0.675, 1.75]
Food Consumed [kg/day]	Fats, Oils, Salad Dressings	−0.2658 [−4.5077, 3.9761]	0.9002	0.767 [0.0110, 53.3]
Food Consumed [kg/day]	Sugars, Sweets, Beverages	−0.0257 [−0.0949, 0.0435]	0.4585	0.975 [0.909, 1.04]

^aSample-weighted participants reporting using cigarettes (and no other tobacco products) five days prior to physical examination and with serum cotinine measurement >10 mg/mL.

^b95% confidence interval.

^cThe dependent variable, biomarker concentration, was natural log-transformed for the regression model.

^dFor each unit-increase in the predictor, the expected biomarker concentration in mg/mL is multiplied by the exponentiated coefficient (controlling for other predictors in the model).

^eRef: reference group.

Table 5.

Sample-weighted multiple linear regression model for urinary 4HBeMA/MHBMA3 concentrations (mg/mL) among 2011–2016 NHANES non-users^a (N=5171).

Predictor	Level	Coefficient [95% CI]b	p-Valuec	Exponentiated slope [95% CI]d
Intercept	Intercept	0.3417 [0.2002, 0.4831]	<.0001	1.41 [1.22, 1.62]
Creatinine, Urine [g/L]	Slope	0.7798 [0.7175, 0.8420]	<.0001	2.18 [2.05, 2.32]
Cotinine, Serum [mg/mL]	Slope	0.0500 [0.0278, 0.0722]	<.0001	1.05 [1.03, 1.07]
Sex	Male	Ref.e	.	Ref.e
Sex	Female	−0.0218 [−0.0903, 0.0467]	0.5257	0.978 [0.914, 1.05]
Age	3–5	0.2985 [0.1372, 0.4597]	0.0005	1.35 [1.15, 1.58]
Age	6–11	0.3697 [0.2703, 0.4690]	<.0001	1.45 [1.31, 1.60]
Age	12–19	−0.0441 [−0.1337, 0.0455]	0.3268	0.957 [0.875, 1.05]
Age	20–39	Ref.e	.	Ref.e
Age	40–59	0.0995 [0.0059, 0.1931]	0.0377	1.10 [1.01, 1.21]
Age	≥60	0.2129 [0.1279, 0.2979]	<.0001	1.24 [1.14, 1.35]
Race/Hispanic Origin	Non-Hispanic White	Ref.e	.	Ref.e
Race/Hispanic Origin	Non-Hispanic Black	−0.0527 [−0.1197, 0.0144]	0.121	0.949 [0.887, 1.01]
Race/Hispanic Origin	Hispanic	0.0078 [−0.0739, 0.0895]	0.8483	1.01 [0.929, 1.09]
Race/Hispanic Origin	Other Race/Multi-Racial	−0.0812 [−0.1904, 0.0280]	0.1413	0.922 [0.827, 1.03]
Weight Status	Underweight	0.1167 [−0.0892, 0.3226]	0.2599	1.12 [0.915, 1.38]
Weight Status	Healthy Weight	Ref.e	.	Ref.e
Weight Status	Overweight/Obesity	−0.0139 [−0.0826, 0.0548]	0.6864	0.986 [0.921, 1.06]
Food Consumed [kg/day]	Milk Products	0.0574 [−0.0542, 0.1690]	0.3062	1.06 [0.947, 1.18]
Food Consumed [kg/day]	Meat, Poultry, Fish	0.0750 [−0.0870, 0.2370]	0.3564	1.08 [0.917, 1.27]
Food Consumed [kg/day]	Eggs	0.1582 [−0.2481, 0.5645]	0.4374	1.17 [0.780, 1.76]
Food Consumed [kg/day]	Legumes, Nuts, Seeds	0.2009 [−0.1161, 0.5179]	0.2085	1.22 [0.890, 1.68]
Food Consumed [kg/day]	Grain Products	0.0248 [−0.0743, 0.1239]	0.6166	1.03 [0.928, 1.13]
Food Consumed [kg/day]	Fruits	0.0159 [−0.1120, 0.1438]	0.8034	1.02 [0.894, 1.15]
Food Consumed [kg/day]	Vegetables	0.0615 [−0.1034, 0.2263]	0.457	1.06 [0.902, 1.25]
Food Consumed [kg/day]	Fats, Oils, Salad Dressings	0.6139 [−1.2757, 2.5036]	0.5166	1.85 [0.279, 12.2]
Food Consumed [kg/day]	Sugars, Sweets, Beverages	−0.0222 [−0.0454, 0.0009]	0.0593	0.978 [0.956, 1.00]

^aSample-weighted participants reporting not using cigarettes (and no other tobacco products) 5 days prior to physical examination and with serum cotinine measurement ≤ 10 mg/mL.

^b95% confidence interval.

^cThe dependent variable, biomarker concentration, was natural log-transformed for the regression model.

^dFor each unit-increase in the predictor, the expected biomarker concentration in mg/mL is multiplied by the exponentiated coefficient (controlling for other predictors in the model).

^eRef: reference group.

Table 6.

Sample-weighted multiple linear regression CPD⁵ model for urinary 4HBeMA/MHBMA3 concentrations (mg/mL) among 2011–2016 NHANES participants (N=5897).

Predictor	Level	Coefficient [95% CI]a	p-Valueb	Exponentiated slope [95% CI]c
Intercept	Intercept	0.380 [0.224, 0.535]	<.0001	1.46 [1.25, 1.71]
Tobacco Smoke Exposure	≤0.015 mg/mL Cotinine, Serum	Refd	.	Refd
Tobacco Smoke Exposure	>0.015 – ≤10 mg/mL Cotinine, Serum	0.0604 [−1.51E-03, 0.122]	0.0556	1.06 [0.998, 1.13]
Tobacco Smoke Exposure	1–10 CPDe	1.75 [1.64, 1.86]	<.0001	5.75 [5.17, 6.39]
Tobacco Smoke Exposure	11–20 CPDe	2.25 [2.08, 2.42]	<.0001	9.49 [7.99, 11.3]
Tobacco Smoke Exposure	>20 CPDe	2.52 [2.20, 2.83]	<.0001	12.4 [9.06, 16.9]
Creatinine, Urine [g/L]b	Slope	0.743 [0.692, 0.794]	<.0001	2.10 [2.00, 2.21]
Sex	Male	Refd	.	Refd
Sex	Female	0.0137 [−0.0521, 0.0794]	0.6776	1.01 [0.949, 1.08]
Age	3–5	0.277 [0.114, 0.440]	0.0013	1.32 [1.12, 1.55]
Age	6–11	0.361 [0.261, 0.460]	<.0001	1.43 [1.30, 1.58]
Age	12–19	−0.0448 [−0.133, 0.0438]	0.3145	0.956 [0.875, 1.04]
Age	20–39	Refd	.	Refd
Age	40–59	0.115 [0.0288, 0.202]	0.0101	1.12 [1.03, 1.22]
Age	≥60	0.214 [0.128, 0.299]	<.0001	1.24 [1.14, 1.35]
Race/Hispanic Origin	Non-Hispanic White	Refd	.	Refd
Race/Hispanic Origin	Non-Hispanic Black	−0.0753 [−0.134, −0.0164]	0.0134	0.927 [0.874, 0.984]
Race/Hispanic Origin	Hispanic	−0.0163 [−0.0918, 0.0593]	0.6673	0.984 [0.912, 1.06]
Race/Hispanic Origin	Other Race/Multi-Racial	−0.0903 [−0.197, 0.0168]	0.0964	0.914 [0.821, 1.02]
Weight Status	Underweight	0.123 [−0.0501, 0.295]	0.16	1.13 [0.951, 1.34]
Weight Status	Healthy Weight	Refd	.	Refd
Weight Status	Overweight/Obesity	−0.0149 [−0.0817, 0.0519]	0.6557	0.985 [0.922, 1.05]
Food Consumed [kg/day]	Milk Products	0.0539 [−0.0458, 0.154]	0.2824	1.06 [0.955, 1.17]
Food Consumed [kg/day]	Meat, Poultry	0.0715 [−0.0636, 0.207]	0.2926	1.07 [0.938, 1.23]
Food Consumed [kg/day]	Eggs	0.214 [−0.176, 0.604]	0.276	1.24 [0.838, 1.83]
Food Consumed [kg/day]	Legumes, Nuts, Seeds	0.174 [−0.122, 0.470]	0.2421	1.19 [0.886, 1.60]
Food Consumed [kg/day]	Grain Products	−7.83E-03 [−0.0949, 0.0793]	0.8572	0.992 [0.909, 1.08]
Food Consumed [kg/day]	Fruits	−0.0271 [−0.144, 0.0899]	0.6429	0.973 [0.866, 1.09]
Food Consumed [kg/day]	Vegetables	0.0210 [−0.156, 0.198]	0.8126	1.02 [0.856, 1.22]
Food Consumed [kg/day]	Fats, Oils, Salad Dressings	0.495 [−1.33, 2.32]	0.5867	1.64 [0.266, 10.1]
Food Consumed [kg/day]	Sugars, Sweets, Beverages	−0.0245 [−0.0495, 4.55E-04]	0.0541	0.976 [0.952, 1.00]

^a95% confidence interval.

^bThe dependent variable, biomarker concentration, was natural log-transformed for the regression model.

^cFor each unit-increase in the predictor, the expected biomarker concentration in mg/mL is multiplied by the exponentiated coefficient (controlling for other predictors in the model).

^dRef: reference group.

^eCPD: cigarettes smoked per day.

Figure 1.

Sample-weighted least-square means [95% confidence intervals] for urinary 4HBeMA/MHBMA3 concentrations (mg/mL) for each CPD category (N=5897).

We used the same three multivariate models to study the association of urinary 34HBMA with tobacco use, diet, and demographic variables. Tobacco smoke exposure was consistently associated with higher urinary 34HBMA compared with non-users in all models, and this effect was dose dependent (Figure 2). Compared with non-users, smoking 1–10, 11–20 and >20 CPD was significantly associated with 32.8%, 44.2%, and 102% higher 34HBMA levels, respectively, controlling for other variables (Table 7). A similar dose-dependent interaction was found when tobacco-smoke exposure was modelled using serum cotinine for both exclusive smokers (p = 0.0476, Table 8) and non-users with SHS exposure (p = 0.0025, Table 9), controlling for other variables.

Figure 2.

Sample-weighted least-square means [95% confidence intervals] for urinary 34HBMA/DHBMA concentrations (mg/mL) for each CPD category (N=5897).

Table 7.

Sample-weighted multiple linear regression CPD⁵ model for urinary 34HBMA/DHBMA concentrations (mg/mL) among 2011–2016 NHANES participants (N=5897).

Predictor	Level	Coefficient [95% CI]a	p-Valueb	Exponentiated slope [95% CI]c
Intercept	Intercept	4.52 [4.41, 4.64]	<.0001	92.0 [82.0, 103]
Tobacco Smoke Exposure	≤0.015 mg/mL Cotinine, Serum	Refd.	.	Refd
Tobacco Smoke Exposure	>0.015 – ≤10 mg/mL Cotinine, Serum	0.0420 [−8.03E-04, 0.0848]	0.0543	1.04 [0.999, 1.09]
Tobacco Smoke Exposure	1–10 CPDe	0.284 [0.217, 0.351]	<.0001	1.33 [1.24, 1.42]
Tobacco Smoke Exposure	11–20 CPDe	0.366 [0.271, 0.462]	<.0001	1.44 [1.31, 1.59]
Tobacco Smoke Exposure	>20 CPDe	0.703 [0.451, 0.955]	<.0001	2.02 [1.57, 2.60]
Creatinine, Urine [g/L]2	Slope	0.822 [0.773, 0.871]	<.0001	2.28 [2.17, 2.39]
Sex	Male	Refd	.	Refd
Sex	Female	0.0179 [−0.0279, 0.0637]	0.435	1.02 [0.973, 1.07]
Age	3–5	0.560 [0.436, 0.684]	<.0001	1.75 [1.55, 1.98]
Age	6–11	0.352 [0.279, 0.425]	<.0001	1.42 [1.32, 1.53]
Age	12–19	0.0514 [5.11E-03, 0.0978]	0.0303	1.05 [1.01, 1.10]
Age	20–39	Refd	.	Refd
Age	40–59	0.0947 [0.0401, 0.149]	0.0011	1.10 [1.04, 1.16]
Age	≥60	0.292 [0.237, 0.346]	<.0001	1.34 [1.27, 1.41]
Race/Hispanic Origin	Non-Hispanic White	Refd	.	Refd
Race/Hispanic Origin	Non-Hispanic Black	−0.145 [−0.195, −0.0962]	<.0001	0.865 [0.823, 0.908]
Race/Hispanic Origin	Hispanic	1.41E-03 [−0.0538, 0.0566]	0.9593	1.00 [0.948, 1.06]
Race/Hispanic Origin	Other Race/Multi-Racial	−0.0999 [−0.166, −0.0341]	0.0037	0.905 [0.847, 0.966]
Weight Status	Underweight	0.107 [−0.0328, 0.246]	0.1308	1.11 [0.968, 1.28]
Weight Status	Healthy Weight	Refd	.	Refd
Weight Status	Overweight/Obesity	−2.46E-03 [−0.0451, 0.0402]	0.9082	0.998 [0.956, 1.04]
Food Consumed [kg/day]	Milk Products	−0.0248 [−0.0884, 0.0387]	0.4355	0.975 [0.915, 1.04]
Food Consumed [kg/day]	Meat, Poultry	−0.0835 [−0.197, 0.0303]	0.1465	0.920 [0.821, 1.03]
Food Consumed [kg/day]	Eggs	7.62E-03 [−0.283, 0.298]	0.9581	1.01 [0.754, 1.35]
Food Consumed [kg/day]	Legumes, Nuts, Seeds	0.0669 [−0.103, 0.237]	0.4326	1.07 [0.902, 1.27]
Food Consumed [kg/day]	Grain Products	0.0839 [6.50E-03, 0.161]	0.0342	1.09 [1.01, 1.17]
Food Consumed [kg/day]	Fruits	0.0341 [−0.0410, 0.109]	0.3658	1.03 [0.960, 1.12]
Food Consumed [kg/day]	Vegetables	2.09E-03 [−0.0966, 0.101]	0.9662	1.00 [0.908, 1.11]
Food Consumed [kg/day]	Fats, Oils, Salad Dressings	0.0818 [−0.910, 1.07]	0.8689	1.09 [0.403, 2.93]
Food Consumed [kg/day]	Sugars, Sweets, Beverages	−0.0189 [−0.0427, 5.01E-03]	0.1186	0.981 [0.958, 1.01]

^a95% confidence interval.

^bThe dependent variable, biomarker concentration, was natural log-transformed for the regression mode.

^cFor each unit-increase in the predictor, the expected biomarker concentration in mg/mL is multiplied by the exponentiated coefficient (controlling for other predictors in the model).

^dRef: reference group.

^eCPD: cigarettes smoked per day.

Table 8.

Sample-weighted multiple linear regression model for urinary 34HBMA/DHBMA concentrations (mg/mL) among 2011–2016 NHANES exclusive smokers^a (N=726).

Predictor	Level	Coefficient [95% CI]b	p-Valuec	Exponentiated slope [95% CI]d
Intercept	Intercept	4.8607 [4.6052, 5.1162]	<.0001	129 [100, 167]
Creatinine, Urine [g/L]	Slope	0.7446 [0.6743, 0.8149]	<.0001	2.11 [1.96, 2.26]
Cotinine, Serum [mg/mL]	Slope	0.0005 [0.0000, 0.0009]	0.0476	1.00 [1.00, 1.00]
Sex	Male	Ref.e	.	Ref.e
Sex	Female	0.0754 [−0.0226, 0.1735]	0.1285	1.08 [0.978, 1.19]
Age	3–5	N/A	.	N/A
Age	6–11	N/A	.	N/A
Age	12–19	−0.0416 [−0.1923, 0.1091]	0.5812	0.959 [0.825, 1.12]
Age	20–39	Ref.e	.	Ref.e
Age	40–59	0.1150 [0.0205, 0.2094]	0.0181	1.12 [1.02, 1.23]
Age	≥60	0.2419 [0.0764, 0.4074]	0.0051	1.27 [1.08, 1.50]
Race/Hispanic Origin	Non-Hispanic White	Ref.e	.	Ref.e
Race/Hispanic Origin	Non-Hispanic Black	−0.1712 [−0.2770, −0.0654]	0.0021	0.843 [0.758, 0.937]
Race/Hispanic Origin	Hispanic	0.0645 [−0.0432, 0.1722]	0.2346	1.07 [0.958, 1.19]
Race/Hispanic Origin	Other Race/Multi-Racial	0.0366 [−0.1820, 0.2552]	0.7377	1.04 [0.834, 1.29]
Weight Status	Underweight	0.0591 [−0.1752, 0.2934]	0.6143	1.06 [0.839, 1.34]
Weight Status	Healthy Weight	Ref.e	.	Ref.e
Weight Status	Overweight/Obesity	−0.0256 [−0.1243, 0.0730]	0.6039	0.975 [0.883, 1.08]
Food Consumed [kg/day]	Milk Products	−0.0652 [−0.2301, 0.0997]	0.4301	0.937 [0.794, 1.10]
Food Consumed [kg/day]	Meat, Poultry, Fish	−0.1175 [−0.3165, 0.0815]	0.2409	0.889 [0.729, 1.08]
Food Consumed [kg/day]	Eggs	−0.0448 [−0.7507, 0.6611]	0.899	0.956 [0.472, 1.94]
Food Consumed [kg/day]	Legumes, Nuts, Seeds	0.2522 [−0.4634, 0.9677]	0.4819	1.29 [0.629, 2.63]
Food Consumed [kg/day]	Grain Products	0.0275 [−0.1275, 0.1825]	0.7226	1.03 [0.880, 1.20]
Food Consumed [kg/day]	Fruits	−0.1793 [−0.5144, 0.1559]	0.2874	0.836 [0.598, 1.17]
Food Consumed [kg/day]	Vegetables	0.1166 [−0.2229, 0.4562]	0.493	1.12 [0.800, 1.58]
Food Consumed [kg/day]	Fats, Oils, Salad Dressings	2.7963 [1.1027, 4.4899]	0.0017	16.4 [3.01, 89.1]
Food Consumed [kg/day]	Sugars, Sweets, Beverages	−0.0297 [−0.0978, 0.0384]	0.3843	0.971 [0.907, 1.04]

^aSample-weighted participants reporting using cigarettes (and no other tobacco products) 5 days prior to physical examination and with serum cotinine measurement >10 mg/mL.

^b95% confidence interval.

^cThe dependent variable, biomarker concentration, was natural log-transformed for the regression model.

^dFor each unit-increase in the predictor, the expected biomarker concentration in mg/mL is multiplied by the exponentiated coefficient (controlling for other predictors in the model).

^eRef: reference group.

Table 9.

Sample-weighted multiple linear regression model for urinary 34HBMA/DHBMA concentrations (mg/mL) among 2011–2016 NHANES non-users^a (N=5171).

Predictor	Level	Coefficient [95% CI]b	p-Valuec	Exponentiated slope [95% CI]d
Intercept	Intercept	4.4990 [4.4079, 4.5902]	<.0001	89.9 [82.1, 98.5]
Creatinine, Urine [g/L]	Slope	0.8394 [0.7815, 0.8974]	<.0001	2.32 [2.18, 2.45]
Cotinine, Serum [mg/mL]	Slope	0.0228 [0.0084, 0.0372]	0.0025	1.02 [1.01, 1.04]
Sex	Male	Ref.e	.	Ref.e
Sex	Female	0.0088 [−0.0373, 0.0548]	0.7033	1.01 [0.963, 1.06]
Age	3–5	0.5862 [0.4689, 0.7034]	<.0001	1.80 [1.60, 2.02]
Age	6–11	0.3707 [0.3040, 0.4373]	<.0001	1.45 [1.36, 1.55]
Age	12–19	0.0671 [0.0229, 0.1113]	0.0037	1.07 [1.02, 1.12]
Age	20–39	Ref.e	.	Ref.e
Age	40–59	0.1013 [0.0426, 0.1600]	0.0011	1.11 [1.04, 1.17]
Age	≥60	0.3083 [0.2498, 0.3669]	<.0001	1.36 [1.28, 1.44]
Race/Hispanic Origin	Non-Hispanic White	Ref.e	.	Ref.e
Race/Hispanic Origin	Non-Hispanic Black	−0.1417 [−0.2077, −0.0756]	<.0001	0.868 [0.812, 0.927]
Race/Hispanic Origin	Hispanic	−0.0051 [−0.0686, 0.0584]	0.873	0.995 [0.934, 1.06]
Race/Hispanic Origin	Other Race/Multi-Racial	−0.1092 [−0.1812, −0.0372]	0.0037	0.897 [0.834, 0.963]
Weight Status	Underweight	0.1296 [−0.0326, 0.2917]	0.1147	1.14 [0.968, 1.34]
Weight Status	Healthy Weight	Ref.e	.	Ref.e
Weight Status	Overweight/Obesity	−0.0001 [−0.0471, 0.0469]	0.9962	1.000 [0.954, 1.05]
Food Consumed [kg/day]	Milk Products	−0.0202 [−0.0940, 0.0536]	0.5848	0.980 [0.910, 1.06]
Food Consumed [kg/day]	Meat, Poultry, Fish	−0.0743 [−0.2004, 0.0518]	0.2419	0.928 [0.818, 1.05]
Food Consumed [kg/day]	Eggs	0.0017 [−0.3162, 0.3196]	0.9917	1.00 [0.729, 1.38]
Food Consumed [kg/day]	Legumes, Nuts, Seeds	0.0660 [−0.1294, 0.2614]	0.5001	1.07 [0.879, 1.30]
Food Consumed [kg/day]	Grain Products	0.0908 [0.0118, 0.1699]	0.0253	1.10 [1.01, 1.19]
Food Consumed [kg/day]	Fruits	0.0694 [0.0007, 0.1380]	0.0479	1.07 [1.00, 1.15]
Food Consumed [kg/day]	Vegetables	−0.0145 [−0.1175, 0.0885]	0.7783	0.986 [0.889, 1.09]
Food Consumed [kg/day]	Fats, Oils, Salad Dressings	−0.3576 [−1.5379, 0.8227]	0.5451	0.699 [0.215, 2.28]
Food Consumed [kg/day]	Sugars, Sweets, Beverages	−0.0128 [−0.0295, 0.0038]	0.127	0.987 [0.971, 1.00]

^aSample-weighted participants reporting not using cigarettes (and no other tobacco products) 5 days prior to physical examination and with serum cotinine measurement ≤10 mg/mL.

^b95% confidence interval.

^cThe dependent variable, biomarker concentration, was natural log-transformed for the regression model.

^dFor each unit-increase in the predictor, the expected biomarker concentration in mg/mL is multiplied by the exponentiated coefficient (controlling for other predictors in the model).

^eRef: reference group.

All regression models predicted that study participants aged 40 years and older had significantly higher 4HBeMA and 34HBMA compared with the reference group (aged 20–39 years); the average percent increase for those aged 40–59 years and >60 years was 3.6%, p ≤ 0.0377 and 1.2%, p ≤ 0.0051, respectively. Among non-users (Table 5) and all study participants (Table 6), all age groups except for adolescents (aged 12–19 years) had higher 4HBeMA levels compared with the reference age group (aged 20–39 years). 34HBMA levels were higher among all age groups for non-users (Table 9) and all study participants (Table 7) compared with the reference age group (aged 20–39 years). The magnitude of this association was much lower than the association with tobacco smoke.

A modest association was also found with race/Hispanic origin. In five of the six models, Non-Hispanic Blacks were found to have significant lower 4HBeMA and 34HBMA levels compared with the reference group (Non-Hispanic Whites) after adjusting for co-variates (p≤0.0134; Tables 4, 8, 9, 6, and 7). The magnitude of this association (7.3–30.9% decrease) was much lower than the association with tobacco smoke. Additionally, Other Race/Multi-Racial ethnicity was associated with decreased 34HBMA (9.5–10.3% decrease; p = 0.0037) in two regression models (Tables 9 and 7).

The 24-h dietary recall of different food groups was rarely and inconsistently associated with urinary 4HBeMA or 34HBMA levels. For 34HBEMA, median grain consumption (0.265 Kg, or 8.9 oz) was associated with 2.44% higher 34HBMA levels (p = 0.0253). Similarly, median fruit consumption (0.067 Kg (2.4 oz)) was associated with 0.47% higher 34HBMA (p = 0.0479).

4. Discussion

Our analysis showed that 1HMPeMA and 2HBeMA were detected at 0.7% and 9.8% in urine specimens, respectively, while 34HBMA and 4HBeMA were both detected in more than 96% of specimens. The low detection rates of 1HMPeMA and 2HBeMA are consistent with previous studies where the concentration of MHBMA (1HMPeMA + 2HBeMA) for smokers and non-smokers was below their LOD (Schettgen et al. 2009, Eckert et al. 2011, Chiang et al. 2015), thus limiting the utility of 1HMPeMA and 2HBeMA as butadiene-exposure biomarkers. Similarly, among MHBMA isomers, 4HBeMA has been detected in urine samples of smokers and non-smokers (St Helen et al. 2014, Chiang et al. 2015). Detection rates for 34HBMA detection rates were high, but the magnitude of the increase attributable to smoking >20 CPD (44%) was relatively low, even though this amount of smoking would lead to inhalation of milligram quantities of 1,3-butadiene (Pazo et al. 2016). The low association with tobacco smoke may indicate the presence of endogenous or exogenous sources of 34HBMA besides tobacco smoke. Conversely, 4HBeMA was widely detected and increased dramatically with increased smoking: an 1143% increase in urinary 4HBeMA was attributable to smoking >20 CPD. Thus, we conclude that 4HBeMA is an effective biomarker of exogenous 1,3-butadiene exposure.

All six of our regression models indicate that cigarette smoke was a major source of 1,3-butadiene exposure in the U.S. population during the 2011–2016 NHANES sampling period. Most significantly, our regression analysis demonstrated that urinary 4HBeMA levels increased with increasing tobacco-smoke exposure as assessed using serum cotinine or CPD categories (Figure 1). The least-square means presented in Figure 1 further show that smoking more CPD was significantly associated with increased urinary 4HBeMA (e.g., 1143% higher 4HBeMA associated with smoking >20 CPD) compared with people who had no tobacco-smoke exposure (serum cotinine ≤0.015 mg/mL) and people presumptively exposed to SHS (serum cotinine >0.015 – ≤10 mg/mL). Additionally, the sample-weighted median 4HBeMA among exclusive-smokers (31.5 μg/g creatinine) was approximately eight times that of non-users (4.11 μg/g creatinine) (Table 2). Similar results were found when cigarette smoke exposure was modelled using serum cotinine instead of CPD: higher smoke exposure was associated with higher urinary 4HBeMA for both active smokers (p < 0.0001, Table 4) and non-users with SHS exposure (p < 0.0001, Table 5).

In the same way, our regression analysis revealed that 34HBMA levels are associated with serum cotinine and follow a dose-dependent pattern with CPD categories (Figure 2), but this increase was less pronounced compared with 4HBeMA. Participants presumptively exposed to SHS (serum cotinine >0.015 – ≤10 mg/mL) had nominally higher 4HBeMA and 34HBMA levels compared with participants who had no tobacco-smoke exposure (serum cotinine ≤0.015 mg/mL), however, this increase was not statistically significant after controlling for confounders in the overall model (Table 6 and 7). Moreover, sample-weighted median 34HBMA among exclusive-smokers (391 μg/g creatinine) was higher than in non-users (296 μg/g creatinine) (Table 3). Interestingly, our analysis model revealed less-significant differences in urinary 34HBMA levels among exclusive smokers, diminishing the utility of 34HBMA as a useful biomarker of smoke exposure.

Together, these results provide evidence of the importance of tobacco smoke as a 1,3-butadiene exposure source. Similar to our findings, previous studies have shown associations of urinary 1,3 butadiene biomarkers with serum cotinine and CPD among smokers (Roethig et al. 2009, Eckert et al. 2011, Chiang et al. 2015, Pluym et al. 2015). Furthermore, our findings clearly demonstrate that 4HBeMA is significantly associated with 1,3-butadiene exposure. The present results are supportive by the recent findings showing that 4HBeMA (alone and reported as the sum of different isomers) is an appropriate biomarker for exposure assessment of carcinogenic 1,3-butadiene from tobacco smoke (De Jesus et al. 2020, Frigerio et al. 2020, Keith et al. 2020). Consistent with our analysis, previous studies indicate that 34HBMA is presumably formed from other exogenous or endogenous sources (Albertini et al. 2003, Carmella et al. 2009, Schettgen et al. 2009, Eckert et al. 2011), hence explaining the less-significant differences in urinary 34HBMA levels among smokers (Figure 2). In one study, for example, 1,3-butadiene biomarker levels were measured before and after cessation of smoking; the authors concluded that 34HBMA was unaffected by smoking status, whereas 4HBeMA was significantly decreased after cessation of smoking (Carmella et al. 2009).

Demographic variables were also evaluated for association with urinary 34HBMA and 4HBeMA in the sample-weighted multiple linear regression model. Our results show higher levels of 34HBMA and 4HBeMA in older adults (aged 40–59 and ≥60 years) compared with younger adults (aged 20–39 years). The higher 34HBMA and 4HBeMA in older adults (≥60 years) could possibly result from endogenous processes related with aging (Lechner et al. 2006). Urinary 4HBeMA levels among children (non-users) (aged 3–5 and 6–11 years) were significantly higher than non-user adults (Table 5). We identified non-users if they answered “no” to “tobacco use within five days prior to NHANES physical examination” or had serum cotinine ≤10 mg/mL. While young children are more likely to be exposed to SHS, especially when living with someone who smokes inside the home (CDC 2010), our models include serum cotinine as a co-variate that would control for SHS exposure. Thus the higher levels found in children (Tables 5 and 9) could be influenced by differences in unassessed exposure or metabolism (Bearer 1995).

Our study also found inconsistent associations of race/ethnicity and gender with urinary 1,3-butadiene metabolites. For example, only one of six models found a gender-specific effect (female smokers have higher 4HBeMA than male smokers, Table 4). Females tend to have lower lean body mass and thus lower creatinine production compared with males (Barr et al. 2005), so their creatinine-ratioed urinary biomarkers may be higher because of a smaller denominator (Table 4). Including creatinine in the regression models as a predictor variable controls for creatinine effects separately and indicates that gender is not consistently predictive of differences in 4HBeMA or 34HBMA. Our report also indicates that being Non-Hispanic Black (compared with Non-Hispanic White) is a negative predictor of urinary 34HBMA and 4HBeMA (five of six models) (Tables 4, 8, 9, 6, and 7). Similarly, being Other Race/Multi-Racial is a negative predictor of urinary 34HBMA in two of the regression models (Tables 9 and 7). Muscle mass tends to be higher in Non-Hispanic Blacks compared to other races, thus resulting in higher levels of urinary creatinine (Barr et al. 2005). Alternatively, this may be the result of differences in how 1,3-butadiene is metabolised, as previously reported (Boldry et al. 2017).

No food groups were consistently found to be significant predictors of urinary 4HBeMA or 34HBMA. Consumption of fats, oils, and salad dressings (one of six models), grain products (two of six models), and fruit (one of six models) were positive predictors of 34HBMA, and the magnitude of increase predicted by the median food intakes were only 9.05%, 2.44%, and 0.47%, respectively. The increase in 34HBMA associated with consumption of fats, oils, and salad dressings is consistent with findings of 1,3-butadiene formation in some heated cooking oils (Shields et al. 1995, Huang et al. 2020). Nonetheless, dietary variables were not consistently associated with butadiene exposure biomarkers, and when they were the predicted effect of median consumption was modest compared with the predicted effect of smoking.

While this study reports significant new data regarding 1,3-butadiene exposure, it was not a comprehensive analysis of this topic. Exposure to 1,3-butadiene, for example, may occur via a number of sources that we did not assess, such as industrial emissions; automobile exhaust, burning of wood, plastics, and rubber (ATSDR 2012); cooking emissions (Huang et al. 2020); and through residual 1,3-butadiene in food packaging (Yurawecz et al. 1976, McNeal and Breder 1987, Abrantes et al. 2000). Moreover, a recent occupation study of the rubber industry reported that urinary metabolites of 1,3-butadiene were higher in workers with known 1,3-butadiene exposure levels (Albertini et al. 2003); we did not include occupation and industry type in our analyses. Lastly, we evaluated exposure based on concentrations of non-persistent biomarkers in a single spot urine specimen instead of a sample collected over a longer time period.

5. Conclusion

Exposure to the volatile organic compound 1,3-butadiene – a human carcinogen – can result in adverse health effects. We provide the first U.S. population-representative report on 1,3-butadiene exposure by assessing its urinary metabolites: 34HBMA, 1HMPeMA, 2HBeMA, and 4HBeMA. This study shows that tobacco smoke is a significant source of 1,3-butadiene exposure in a representative sample of the U.S. population. Our data also indicate that 4HBeMA is a suitable biomarker for assessing non-occupational 1,3-butadiene exposure. Biomonitoring of urinary 4HBeMA in subsequent NHANES cycles will provide useful data about changes in 1,3-butadiene exposure in the U.S. population.