Evaluation of Tobacco Smoke and Diet as Sources of Exposure to Two Heterocyclic Aromatic Amines for the U.S. Population: NHANES 2013–2014

Abstract

Background:

Heterocyclic aromatic amines (HAAs) are a group of hazardous substances produced during combustion of tobacco or high-temperature cooking of meats. 2-Amino-9H-pyrido[2,3-b]indole (AαC) is a major carcinogenic HAAs in tobacco smoke.

Methods:

Urinary AαC, used as a marker of AαC exposure, was analyzed on spot urine samples from adult participants of the 2013-2014 cycle of the National Health and Nutrition Examination Survey (NHANES; N=1,792). AαC was measured using isotope-dilution liquid chromatography–tandem mass spectrometry. Exclusive combusted tobacco smokers were differentiated from non-users of tobacco products through both self-report and serum cotinine data.

Results:

Among exclusive smokers, sample-weighted median urinary AαC was 40 times higher than non-users. Sample-weighted regression models showed that urinary AαC increased significantly with serum cotinine among both exclusive tobacco users and non-users with second-hand smoke exposure. Among non-users, eating beef cooked at high temperature was associated with a significant increase in urinary AαC, while consuming vegetables was associated with decreased AαC. In addition, smoking one-half pack of cigarettes per day was associated with a significant increase of 23.6 pg AαC/mL calculated at geometric mean of AαC, controlling for potential confounders. In comparison, increase in AαC attributable to consuming the 99^th percentile of beef cooked at high temperature was 0.99 pg AαC/mL.

Conclusions:

Both exclusive smokers and non-users of tobacco in the general U.S. population are exposed to AαC from tobacco smoke, with additional, lesser contributions from certain dietary components.

Impact:

AαC is an important biomarker that is associated with tobacco smoke exposure.

Introduction

Tobacco use is the single largest preventable cause of disease and death in the United States, and tobacco smoking is responsible for more than 480,000 deaths each year in the United States (1). A significant body of evidence accumulated over several decades suggests that tobacco smoke exposure is related to development of lung cancer and other cancers (1–4). Various carcinogens have been identified in tobacco smoke (4–6).

Heterocyclic aromatic amines (HAAs) are a group of hazardous substances produced during burning of tobacco or high-temperature cooking of meats (7–10). HAAs are amines that contain at least one heterocyclic ring and one aromatic ring (Supplementary Figure S1). HAAs are strongly mutagenic and carcinogenic in various in vitro and in vivo models. HAAs can induce tumors in various organs, including mammary glands, prostate, lungs, colon, skin, bladder, and liver (11–14). Epidemiological studies associate frequent HAA exposure with elevated cancer risk (15–17). The International Agency for Research on Cancer (IARC) categorized a number of HAAs as possible (Group 2B) and probable (Group 2A) human carcinogens (18).

Among more than 25 HAAs so far identified (19), 2-amino-9H-pyrido[2,3-b]indole (AαC, Group 2B carcinogen) is one of most abundant carcinogenic HAAs in tobacco smoke (10). AαC levels in tobacco smoke can reach as high as 260 ng per cigarette (7,10,20,21). This approaches the levels of other well-known carcinogens in tobacco smoke, such as N’-nitrosonornicotine (NNN), 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), and benzo[α]pyrene (BaP) (5,22,23).

Another carcinogenic HAA detected in tobacco smoke is 2-amino-3-methyl-9H-pyrido[2,3-b]indole (MeAαC, Group 2B carcinogen) (Supplementary Figure S1). MeAαC is a methyl homolog of AαC. MeAαC levels in tobacco smoke are about 10-fold lower than AαC levels (7,8,10). These two HAAs are carcinogenic in animal models (5,24–28), and IARC classifies AαC and MeAαC as suspected human carcinogens (29). Additionally, the U.S. Food and Drug Administration (FDA) lists AαC and MeAαC as harmful and potentially harmful constituents (HPHCs) in tobacco products and tobacco smoke (30). Besides tobacco smoke, AαC and MeAαC can form in foods prepared at high temperatures (approximately 200°C–300°C), such as barbecued, fried, or broiled meats, poultry, and fish (8,31–34).

Metabolic activation has been proposed as a mechanism for the carcinogenicity of AαC and MeAαC (9,35). Cytochrome P450 enzymes are involved in the oxidation of the exocyclic amine group of AαC and MeAαC (35–37). Sequentially, these N²-hydroxylated metabolites of AαC and MeAαC can be further catalyzed by acetyltransferases, sulfotransferases, or glucuronosyltransferases to generate reactive O-esters that could bind covalently to DNA and elicit genotoxicity (38–43).

Although the potential for AαC and MeAαC to harm human health is well documented, few studies have evaluated human exposure to these two carcinogens. A small-scale study (170 subjects) conducted in China documented that AαC was consistently detected in smokers’ urine. The number of cigarettes smoked per day was positively associated with urinary levels of AαC in study participants (44). A U.S. study (30 subjects) detected AαC in most of the urine samples of smokers, but those levels had dropped 87% after 6 weeks of not smoking (45). Even fewer studies have measured MeAαC in human biospecimens. The aim of this report is to characterize human exposure to AαC and MeAαC in the general population. We measured AαC and MeAαC as urinary biomarkers of exposure to these carcinogenic HAAs as part of the National Health and Nutrition Examination Survey (NHANES), which conducts biomonitoring of the United States civilian, non-institutionalized population.

Materials and Methods

Study Design

NHANES is a national survey that assesses population health and nutritional status based on data collected from a cross-sectional, multistage probability sample that is representative of the non-institutionalized U.S. civilian population. As part of Centers for Disease Control and Prevention (CDC), the National Center for Health Statistics (NCHS) is responsible for conducting NHANES and collecting questionnaire data, physical examination data, and biospecimens from participants. The NCHS Research Ethics Review Board reviewed and approved the study, and informed written consent was obtained from all participants before they took part in the study.

Spot urine samples were collected during physical examinations carried out in a mobile examination center (MEC) from participants aged ≥18 years in the NHANES 2013–2014 survey cycle. We subsequently measured urinary AαC and MeAαC in a special one-third sample of MEC participants (NHANES dataset HCAAS_H; N = 2,605) that included all adult smokers in the 2013–2014 cycle. Results reported here, however, are from a subset of these participants who met eligibility criteria and for whom all required data were present (this attrition is detailed in Identifying Users of Combusted Tobacco and Non-Users of Tobacco below).

Analytical Method

Spot urine samples were stored at −70°C before assay. The total urinary concentrations of AαC and MeAαC (free and conjugated forms) were analyzed using isotope-dilution liquid chromatography–tandem mass spectrometry (LC/MS/MS) and robotic sample preparation (46). Briefly, an aliquot of 0.5 mL of urine samples was spiked with internal standards that are stable isotope analogs of targeted analytes. The conjugated forms of AαC and MeAαC in the samples were hydrolyzed at base condition for 5 hr. After hydrolysis, the total urinary AαC and MeAαC were extracted from urine matrix by using diatomaceous earth and mixed mode of cation exchange solid phase extraction plates, respectively. The throughput and precision of sample preparation was improved through automated solid phase extraction using an integrated robotic system (46). Analytes were chromatographically resolved from urinary interferences using a reversed phase column (Agilent Zorbax Eclipse Plus C18, 2.1×100 mm 3.5 μm; Santa Clara, CA, USA). AαC and MeAαC in the samples were detected by AB Sciex API 5500 QTRAP system (Framingham, MA, USA). The limits of detection of AαC and MeAαC were 0.62 pg/mL and 0.33 pg/mL, respectively. Serum cotinine was measured by using LC/MS/MS (47).

Low-concentration and high-concentration quality control materials and blank urines were run together with NHANES 2013–2014 samples to evaluate method performance on the day of analysis. The reported data satisfy the accuracy and precision requirements of the quality control/quality assurance program of the CDC National Center for Environmental Health, Division of Laboratory Sciences (48). Measurements below the limit-of-detection were substituted with the limit of detection divided by the square-root of two (49,50).

Identifying Users of Combusted Tobacco and Non-Users of Tobacco

Exclusive combusted tobacco smokers were differentiated from non-users of tobacco products through both self-report and serum cotinine data (51,52), which is detailed in Supplementary Information. Attrition of participants for statistical analysis is as follows: missing serum cotinine data (146 participants), use of smokeless tobacco and nicotine replacement therapy (135 participants), or missing data for other variables involved in regression models (532 participants). This attrition resulted in 1,792 study participants eligible for statistical analysis for AαC and 1,793 study participants for MeAαC. For the cigarettes smoked per day (CPD) regression model described below, 32 additional participants were excluded for missing CPD data leaving 1,760 study participants.

Statistical Analysis

NHANES recruits participants by using a multistage, probability sampling design. This complex design must be accounted for in order to estimate variances correctly and to achieve unbiased, nationally representative statistics. Robust estimation can be made by implementing survey sample weights (NHANES Special Sample weight; WTFSM) on each participant’s data and performing Taylor series linearization. This estimation approach was conducted in the statistical software applications SUDAAN®, Version 11.0.0 (Research Triangle Institute, Research Triangle Park, NC) and SAS® 9.4 (SAS Institute Inc. Cary, NC). Data from the NHANES 2013 – 2014 sampling cycle were analyzed with sample-weighted linear regression models that was stratified by tobacco use status (exclusive smokers vs. non-users). Parameters, including absolute change in biomarker concentration (ΔY) relative to the jth predictor ΔX_j, from these models were estimated as described in detail in Supplementary Information.

Sample-weighted regression models were stratified by smoking status, and a collection of demographic variables as follows were included as predictors: sex, age, race/Hispanic origin, body mass index (BMI), impoverishment (poverty income ratio < 1.00, indicating self-reported family income below the U.S. Census poverty threshold), and fasting time (delta time between the time of specimen collection and last consumption of anything other than water). Except for BMI, information for these potential confounders was self-reported. The reference group was male for sex and non-Hispanic white for race/Hispanic origin. Age in years was divided into ranges: 18–39, 40–59, and ≥60, with 40–59 years as the reference group. For adults ≥20 years BMI was defined using standardized cut-points: underweight (BMI < 18.5), healthy weight (18.5 ≤ BMI < 25), and overweight/obese (BMI ≥ 25). Participants < 20 years were identified as underweight, healthy weight, and overweight/obese if they were below the 5th percentile, between the 5th and 85th percentile, and above the 85th percentile, respectively, for their sex and age (www.cdc.gov/healthyweight/assessing/bmi/adult_bmi/index.html).

Cotinine is a metabolite of nicotine, the primary addictive ingredient in tobacco products. Cotinine in human serum is a highly specific biomarker of tobacco smoke exposure. With a half-life of 16–18 hours, cotinine is suitable for biomarker studies concerning recent tobacco use (52). Serum cotinine was included as a continuous predictor to reflect tobacco smoke exposure. Among tobacco non-users, tobacco smoke exposure is attributable to secondhand tobacco smoke (SHS). SHS exposure can be assessed based on serum cotinine concentrations. In addition, we tested direct association of urinary biomarker concentrations with frequency of tobacco smoking. To run this test, unstratified, sample-weighted regression models where serum cotinine was replaced with a measure of exposure comprising self-reported average number of cigarettes smoked per day (CPD) over the five days preceding the NHANES physical exam. Tobacco smoke exposure in the CPD models was divided into five groups. Group 1 (unexposed to tobacco smoke) included participants with the ranges of ≤0.05 ng/mL serum cotinine; Group 2 (presumptively exposed to second-hand tobacco smoke) included participants with the range of >0.05 to ≤10 ng/mL serum cotinine; Group 3, 4, 5 included participants who consumed 1–10 CPD (0.5 pack), 11–20 (1 pack), and >20 (>1 pack), respectively. The group of unexposed participants was assigned as the reference category. The unexposed category was designated as ≤0.05 ng/mL serum cotinine, which was its limit of detection in the 1999–2000 NHANES cycle. To allow historical comparison of serum cotinine results, we kept using the limit of detection at 0.05 ng/mL for definition of unexposed participants, though this was improved in 2001 to 0.015 ng/mL.

Exposure through diet is another potential source of AαC, especially when food is prepared at high temperature (31,33,34). Dietary exposure was assessed based on the amount participants consumed within each U.S. Department of Agriculture (USDA) food group for the 24-hr period (midnight to midnight) (51), which is described in detail in Supplementary Information. For this assessment, two additional food subgroups were distinguished because of their potential for high AαC exposure arising from high temperature (i.e., broiling, baking, or frying) during preparation (34): high-temperature cooked beef and high-temperature cooked fish. Double counting was avoided by subtracting the amount consumed in each subgroup from the amount consumed in their respective food group. Supplementary Table S1 details the USDA food codes and logic for apportioning dietary intake.

Quantitative analysis of urinary biomarker concentration data must account for normal physiologic variations in urine dilution, which can vary markedly for an individual between voids and thereby confound statistical inference (53). One approach to minimize this problem is to normalize the urinary concentrations of the biomarker to creatinine, which is formed endogenously by lean muscle tissue and cleared by the kidneys into urine at a relatively constant rate. Summary statistics of urinary concentrations are reported as the ratio of AαC and MeAαC to creatinine (ng/g creatinine), and urinary creatinine concentration is included as a continuous predictor in regression models to adjust for potential confounding from urine dilution.

In order to compare the exposure through diet and smoking, we estimated the absolute change in urinary HAA (ΔY) predicted for these two routes under conservative assumptions of exposure. For diet, we estimated the change in urinary HAA in pg/mL associated with consuming the sample-weighted 99^th percentile daily amount of specific food groups in the United States. This dietary exposure is then readily compared to the change in urinary HAA, also in units of pg/mL, associated with smoking one-half pack of cigarettes per day from the CPD model.

Results

Sample-weighted demographic distributions of NHANES 2013–2014 participants (AαC: N = 1,792, MeAαC: N = 1,793) were summarized in Table 1. Among these participants, AαC was detected in 61.5 percent of participants, but only in 30.6 percent for MeAαC, so MeAαC was excluded from multiple regression analysis. By comparison, AαC and MeAαC were detected at much higher rates (98.9% and 75.6%, respectively) in exclusive tobacco smokers than those in non-users (39.5 % for AαC and 4.2% for MeAαC). Among exclusive smokers, urinary AαC and MeAαC concentrations were significantly correlated, with a sample-weighted Pearson correlation of 0.788. The sample-weighted median of urinary AαC among exclusive smokers was 40 times higher than non-users (31.90 vs. 0.79 ng/g creatinine, respectively; Table 2). Geometric means and selected percentiles of urinary AαC concentrations for U.S. population were summarized in Supplementary Table S2 and S3. The difference in urinary concentrations of MeAαC in exclusive tobacco smokers and non-users at selected percentiles was summarized in Supplementary Table S4 and S5. The percent-distribution of urinary AαC and MeAαC depicted these differences among exclusive smokers and non-users (Figure 1).

Table 1.

Sample-weighted demographic proportions for AαC (N = 1,792) and MeAαC (N = 1,793) among NHANES 2013–2014 participants.

	AαC 1				MeAαC 1
Variables	Sample Size 2	Percent (SE 3)	Sample Size 2	Percent (SE 3)	Sample Size 2	Percent (SE 3)	Sample Size 2	Percent (SE 3)
	Tobacco smokers		Non-tobacco users		Tobacco smokers		Non-tobacco users
Age (years)
18–39	276	46.7 (3.08)	419	36.0 (1.43)	276	46.7 (3.01)	419	36.1 (1.41)
40–59	264	40.1 (3.17)	347	34.4 (1.78)	264	40.0 (3.03)	347	34.5 (1.75)
≥60	123	13.3 (1.59)	363	29.6 (1.79)	124	13.3 (1.57)	363	29.4 (1.72)
BMI
Healthy	222	34.9 (1.79)	337	30.1 (1.86)	223	35.0 (1.81)	337	30.0 (1.88)
Over/Obese	421	62.4 (2.02)	779	69.2 (1.80)	420	62.3 (2.06)	779	69.3 (1.82)
Under	20	2.70 (0.65)	13	0.77 (0.17)	21	2.73 (0.64)	13	0.77 (0.17)
Poverty Income Ratio
PIR<1	260	28.3 (3.61)	213	11.8 (1.55)	261	28.3 (3.42)	213	11.8 (1.54)
PIR>=1	403	71.7 (3.61)	916	88.2 (1.55)	403	71.7 (3.42)	916	88.2 (1.54)
Race
Mexican American	43	6.21 (2.25)	190	10.4 (2.21)	43	6.21 (2.26)	190	10.4 (2.21)
Non-Hispanic Black	172	17.0 (2.49)	185	9.13 (1.39)	172	17.1 (2.45)	185	9.15 (1.39)
Non-Hispanic White	349	67.4 (3.85)	478	66.9 (3.70)	351	67.4 (3.80)	478	66.9 (3.67)
Other Hispanic	38	3.52 (1.32)	113	5.52 (0.96)	37	3.46 (1.26)	113	5.54 (0.96)
Other/Multi Race	61	5.84 (1.11)	163	8.03 (0.85)	61	5.85 (1.11)	163	8.05 (0.85)
Sex
Female	301	45.6 (2.31)	606	52.8 (1.78)	303	45.9 (2.42)	605	52.7 (1.79)
Male	362	54.4 (2.31)	523	47.2 (1.78)	361	54.1 (2.42)	524	47.3 (1.79)

¹Same data as in stratified serum cotinine regression models.

²Sample size, unweighted.

³Standard error.

Table 2.

Sample-weighted median (25th, 75th percentile) of urinary AαC (N = 1,792) and MeAαC (N = 1,793) concentrations (creatinine adjusted) by age, sex, body mass index (BMI), and race/Hispanic origin, categorized by smoking status among U.S. population.

	AαC 1		MeAαC 1
Variables	[ng/g creatinine (25th, 75th percentile)]		[ng/g creatinine (25th, 75th percentile)]
	Tobacco smokers §	Non-tobacco users ¶	Tobacco smokers ‡	Non-tobacco users ⁑
All	31.9 [10.2, 79.4]	0.79 [0.45, 1.51]	1.57 [0.35, 4.43]	0.26 [0.16, 0.49]
Age (years)
18–39	19.7 [5.64, 51.7]	0.79 [0.45, 1.64]	0.90 [0.23, 2.85]	0.23 [0.15, 0.45]
40–59	46.0 [18.5, 86.7]	0.83 [0.43, 1.52]	2.28 [0.53, 5.22]	0.29 [0.17, 0.53]
≥60	49.5 [18.6, 105.2]	0.77 [0.46, 1.29]	2.44 [0.66, 6.66]	0.28 [0.18, 0.52]
BMI
Healthy	35.4 [8.87, 74.27]	0.93 [0.54, 1.69]	1.84 [0.33, 4.27]	0.37 [0.20, 0.55]
Over/Obese	30.2 [9.18, 73.3]	0.72 [0.42, 1.35]	1.40 [0.35, 3.96]	0.24 [0.16, 0.46]
Under	84.8 [16.6, 108]	1.41 [0.49, 2.12]	5.60 [1.02, 7.78]	0.68 [0.23, 1.11]
Poverty Income Ratio
PIR<1	40.6 [15.4, 88.7]	0.69 [0.40, 1.34]	2.17 [0.51, 5.26]	0.22 [0.15, 0.38]
PIR>=1	29.2 [6.68, 71.0]	0.80 [0.46, 1.52]	1.47 [0.30, 4.21]	0.27 [0.17, 0.51]
Race
Mexican American	5.44 [1.84, 15.71]	0.83 [0.48, 1.62]	0.29 [0.12, 0.53]	0.25 [0.16, 0.48]
Non-Hispanic Black	27.4[9.24, 50.0]	0.44 [0.31, 0.86]	1.19 [0.32, 2.96]	0.18 [0.13, 0.24]
Non-Hispanic White	38.1 [14.8, 92.4]	0.83 [0.47, 1.55]	1.73 [0.44, 5.27]	0.28 [0.17, 0.52]
Other Hispanic	37.3 [23.3, 67.7]	0.75 [0.47, 1.52]	2.12 [1.17, 3.47]	0.25 [0.15, 0.39]
Other/Multi Race	33.4 [10.8, 51.4]	0.79 [0.55, 1.62]	1.38 [0.35, 3.80]	0.37 [0.21, 0.60]
Sex
Female	42.5 [15.1, 91.6]	0.86 [0.52, 1.75]	1.94 [0.41, 5.53]	0.34 [0.19, 0.59]
Male	26.3 [6.55, 58.0]	0.70 [0.39, 1.29]	1.47 [0.33, 3.54]	0.21 [0.14, 0.40]

¹Same data as in stratified serum cotinine regression models.

Detection rates: §: 98.2%; ¶: 39.3%; ‡:74.8%; ⁑: 4.3%.

Figure 1.

Histograms of urinary AαC and MeAαC (ng analyte/g creatinine) among exclusive tobacco smokers and non-users in the NHANES 2013–2014. Green reference line represents median; A panel showed the percent-distribution of urinary AαC among non-users; B panel showed the percent-distribution of urinary AαC among exclusive tobacco smokers; C panel showed the percent-distribution of urinary MeAαC among non-users; D panel showed the percent-distribution of urinary MeAαC among exclusive tobacco smokers.

The sample-weighted multiple regression model among exclusive smokers (Table 3) found that urinary AαC was positively associated with serum cotinine (0.113 pg AαC/mL per ng cotinine/mL), controlling for sex, age, race/Hispanic origin, BMI, fasting time, urinary creatinine, diet, and impoverishment. Mexican-American exclusive smokers had significantly lower AαC in comparison with non-Hispanic White exclusive smokers. Urinary AαC level among impoverished exclusive smokers was significantly higher than among non-impoverished participants.

Table 3.

Sample-weighted multiple regression results for urinary AαC concentrations [pg/mL] among exclusive smokers from NHANES 2013–2014 participants. The geometric mean of urinary AαC among exclusive smokers used for computing ΔY is 22.6 pg/mL.

Variables	Exclusive Smokers (N = 663)
	Slope [95% CI]1	ΔY [95% CI]2	p–value
Cotinine, serum (ng/mL)	0.0050 [0.0039, 0.0061]	0.113 [0.0900, 0.136]	<.0001
Creatinine, urine (g/mL)	293 [43.0, 543]	4.37E+128 [5.78E+28, 3.31E+228]	0.03
Fasting time (hours)	−0.0227 [−0.0510, 0.0056]	−0.507 [−1.07, 0.0747]	0.11
Age (years)
18–39	−0.149 [−0.497, 0.199]	−3.13 [−8.45, 4.19]	0.37
40–59	Ref.		.
≥60	0.250 [−0.187, 0.687]	6.40 [−3.18, 20.7]	0.24
Food Group
High temperature cooked beef (kg)	0.836 [−1.29, 2.96]	29.5 [−15.2, 345]	0.42
High temperature cooked fish (kg)	0.162 [−0.580, 0.905]	3.98 [−9.15, 30.0]	0.65
Meat, poultry, fish, and mixtures (kg)	0.0157 [−0.565, 0.597]	0.358 [−9.13, 16.5]	0.96
Milk and milk products (kg)	−0.152 [−0.586, 0.283]	−3.18 [−9.56, 6.35]	0.47
Eggs (kg)	0.893 [−1.23, 3.01]	32.5 [−14.7, 365]	0.38
Legumes, nuts, seeds (kg)	−0.922 [−3.28, 1.44]	−13.6 [−21.5, 56.0]	0.42
Grain products (kg)	−0.180 [−0.471, 0.112]	−3.71 [−8.14, 2.09]	0.21
Fruits (kg)	−0.0643 [−0.449, 0.321]	−1.40 [−7.71, 7.58]	0.73
Vegetables (kg)	−0.146 [−0.987, 0.694]	−3.07 [−13.6, 19.7]	0.72
Fats, oils, salad dressings (kg)	0.197 [−4.60, 4.99]	4.91 [−22.2, 2.24E+03]	0.93
Sugars, sweets, beverages (kg)	0.0578 [−0.0206, 0.136]	1.34 [−0.321, 3.13]	0.14
BMI
Healthy	Ref.		.
Overweight/Obese	0.140 [−0.186, 0.467]	3.40 [−3.34, 12.5]	0.37
Underweight	−0.116 [−0.930, 0.698]	−2.47 [−13.1, 19.9]	0.77
Poverty income ratio (PIR)
PIR ≥1	Ref.		.
PIR <1	0.273 [0.0584, 0.487]	7.08 [1.77, 13.5]	0.02
Race
Mexican-American	−0.588 [−1.08, −0.0938]	−10.0 [−14.6, −2.82]	0.02
Non-Hispanic black	−0.189 [−0.545, 0.167]	−3.88 [−9.09, 3.35]	0.28
Non-Hispanic white	Ref.		.
Other Hispanic	0.212 [−0.379, 0.803]	5.33 [−6.37, 25.5]	0.46
Other/Multi-race	−0.173 [−0.563, 0.217]	−3.58 [−9.30, 4.60]	0.36
Sex
Female	0.221 [−0.0862, 0.530]	5.60 [−1.34, 14.8]	0.15
Male	Ref.		.

Abbreviations: Ref, reference group.

^1.The dependent variable, biomarker concentration, was natural log-transformed for the regression model.

^2.ΔY is the expected change in biomarker concentration in pg/mL associated with a unit-increase in the predictor, controlling for other predictors in the model and calculated at the overall geometric mean.

Among non-users of tobacco products (Table 4), urinary AαC was positively associated with serum cotinine (0.140 pg AαC/mL per ng cotinine/mL), presumptively attributable to secondhand smoke exposure, controlling for sex, age, race/Hispanic origin, BMI, fasting time, urinary creatinine, diet, and impoverishment. Non-Hispanic Black and Other/Multi-Racial non-users had lower AαC than non-Hispanic White non-users. Urinary AαC was also positively associated with the amount consumed (kg) of beef cooked under high temperature (15.5 pg/mL per kg), whereas the association with consumption of fish cooked at high temperature was not significant. Consumption of vegetables (−0.264 pg/mL per kg) was a significant negative predictor of urinary AαC. Fasting time also was negatively associated with AαC.

Table 4.

Sample-weighted multiple regression results for urinary AαC concentrations [pg/mL] among non-users of tobacco from NHANES 2013–2014 participants. The geometric mean of urinary AαC among non-users used for computing ΔY is 0.75pg/mL.

Variables	Non–Users (N = 1,129)
	Slope [95% CI]1	ΔY [95% CI]2	p–value
Cotinine, serum (ng/mL)	0.171 [0.0771, 0.264]	0.140 [0.0665, 0.220]	0.001
Creatinine, urine (g/mL)	304 [193, 415]	8.13E+131 [4.58E+87, 1.45E+176]	<.0001
Fasting time (hours)	−0.0161 [−0.0256, −0.0066]	−0.0120 [−0.0185, −5.53E-03]	0.003
Age (years)
18–39	0.0301 [−0.148, 0.208]	0.0230 [−0.0940, 0.161]	0.72
40–59	Ref.	Ref.	.
≥60	−0.128 [−0.293, 0.0375]	−0.0903 [−0.184, 0.0184]	0.12
Food Group
High temperature cooked beef (kg)	3.07 [0.337, 5.81]	15.5 [0.561, 201]	0.03
High temperature cooked fish (kg)	−0.501 [−1.07, 0.0736]	−0.296 [−0.484, 0.0209]	0.08
Meat, poultry, fish, and mixtures (kg)	0.245 [−0.155, 0.645]	0.209 [−0.0870, 0.637]	0.21
Milk and milk products (kg)	−0.0927 [−0.298, 0.112]	−0.0667 [−0.185, 0.0757]	0.35
Eggs (kg)	−0.200 [−1.06, 0.662]	−0.137 [−0.474, 0.609]	0.63
Legumes, nuts, seeds (kg)	−0.0525 [−0.892, 0.787]	−0.0385 [−0.423, 0.793]	0.90
Grain products (kg)	−0.0978 [−0.417, 0.222]	−0.0702 [−0.244, 0.163]	0.52
Fruits (kg)	−0.0182 [−0.271, 0.235]	−0.0136 [−0.167, 0.180]	0.88
Vegetables (kg)	−0.431 [−0.775, −0.0863]	−0.264 [−0.396, −0.0812]	0.02
Fats, oils, salad dressings (kg)	1.198 [−0.838, 3.23]	1.74 [−0.369, 15.5]	0.23
Sugars, sweets, beverages (kg)	0.0236 [−0.0170, 0.0642]	0.0180 [−0.0102, 0.0473]	0.23
BMI
Healthy	Ref.		.
Overweight/Obese	−0.0304 [−0.204, 0.144]	−0.0225 [−0.131, 0.104]	0.72
Underweight	−0.0487 [−0.500, 0.403]	−0.0358 [−0.280, 0.333]	0.82
Poverty income ratio (PIR)
PIR ≥1	Ref.		.
PIR <1	−0.0502 [−0.206, 0.105]	−0.0368 [−0.132, 0.0731]	0.50
Race
Mexican-American	0.0523 [−0.189, 0.294]	0.0404 [−0.118, 0.238]	0.65
Non-Hispanic black	−0.325 [−0.520, −0.130]	−0.209 [−0.298, −0.102]	0.003
Non-Hispanic white	Ref.		.
Other Hispanic	−0.0235 [−0.253, 0.206]	−0.0175 [−0.157, 0.155]	0.83
Other/Multi-race	−0.147 [−0.280, −0.0141]	−0.103 [−0.178, −0.0184]	0.03
Sex
Female	−0.0271 [−0.106 0.0513]	−0.0202 [−0.0712, 0.0347]	0.47
Male	Ref.		.

Abbreviations: Ref, reference group.

^1.The dependent variable, biomarker concentration, was natural log-transformed for the regression model.

^2.ΔY is the expected change in biomarker concentration in pg/mL associated with a unit-increase in the predictor, controlling for other predictors in the model and calculated at the overall geometric mean.

In the un-stratified, sample-weighted CPD regression model for non-users and exclusive smokers (Supplementary Table S6), each CPD exposure group had significantly higher urinary AαC levels compared with those of non-users whose serum cotinine levels were <0.05 ng/mL, controlling for sex, age, race/Hispanic origin, BMI, fasting time, urinary creatinine, diet, and impoverishment. Figure 2 displays the sample-weighted least-square means of urinary AαC from the CPD model by exposure category, where it can be seen that urinary AαC increased in a dose-dependent manner across exposure categories.

Figure 2.

Least-square means of urinary AαC concentrations [pg/mL] , adjusted for sex, age, race/Hispanic origin, BMI, fasting time, urinary creatinine, diet, and impoverishment (N = 1,760). (Error bars denote 95% confidence intervals). Group ⩽0.05 ng/mL serum cotinine: participants unexposed to tobacco smoke; Group >0.05 to ⩽10 ng/mL serum cotinine: participants presumptively exposed to second-hand tobacco smoke; Exclusive tobacco smokers were further classified based on their cigarettes smoked per day (CPD) consumed.

Discussion

From a national, population-based study, we show that tobacco smoke and diet are significantly associated with AαC exposure. Sample-weighted linear regression models showed that serum cotinine was a significant positive predictor of urinary AαC among both exclusive smokers and non-users. Moreover, we found a significant dose-dependent association between exclusive smokers’ urinary AαC levels and CPD smoked. Our findings for tobacco smoke exposure among both exclusive smokers and non-users is consistent with the fact that AαC is the most abundant carcinogenic HAAs in tobacco smoke (7,10,20,21), and indicated that tobacco smoke is a major source of AαC exposure in the U.S. population.

Only three small-scale studies previously reported human AαC exposure levels. A study in China (78 subjects) found that the mean of total urinary concentrations of AαC (free and conjugated forms) for smokers (20 ng/g creatinine) was significantly higher than that for nonsmokers (7 ng/g creatinine) (54). Compared with above study, median urinary AαC was slightly higher for exclusive smokers (31.90 ng/g creatinine) and much lower for non-users (0.79 ng/g creatinine) in the NHANES 2013–2014 samples. The lower median urinary AαC of non-user in our study could be related to lower levels of second-hand smoke exposure in the U.S. than in China. Another study in China measured free urinary AαC in 170 volunteers (44). The mean urinary AαC among light (1–19 CPD) smokers was 7.50 ng/g creatinine (9.2 pg/mL urine) and among heavy (>20 CPD) smokers it was 11.92 ng/g creatinine (13.8 pg/mL urine), while mean urinary AαC among nonsmokers was 2.54 ng/g creatinine. In contrast to Turesky et al. 2007, we observed higher exposure levels to AαC at corresponding CPD (least squared means 1-10 CPD: 14.0 pg/mL; 11-20 CPD: 37.2 pg/mL; >20 CPD: 64.1 pg/mL). These differences may be explained, at least in part, to differences in sample preparation. Turesky et al. 2007 measured free AαC in non-hydrolyzed samples, whereas our measurements were preceded by hydrolysis, which enables measurement of both free and base-sensitive conjugated species of AαC, thereby leading to systematically higher measured concentrations. In addition, there is large variation of AαC in cigarettes between manufacturers that would increase the observed variation in urinary AαC at the same CPD consumed in the two countries (Manabe et al. 1991; Smith et al. 2004; Zhang et al. 2011). Nonetheless, both studies corroborated that urinary AαC followed a dose response pattern with CPD. In one small U.S. study (30 subjects) where urinary AαC measurements comprised both free and base-sensitive conjugated species (45), the measurements had magnitudes closer to ours. Konorev et al. 2015 observed that mean urinary AαC among smokers decreased from 24.3 ng/g creatinine to 3.2 ng/g creatinine after subjects had stopped smoking tobacco for six weeks.

In the NHANES 2013-2014 population, urinary AαC in non-Hispanic White smokers was around 7-fold higher than that in Mexican American smokers (Table 2).The effect of race/Hispanic origin cannot be completely explained by differences in tobacco smoke exposure alone because serum cotinine was included in the model. Difference in toxicokinetics may partly contribute to the race/Hispanic origin-related differences. CYP1A2 is one of the major P450 isoforms responsible for AαC metabolism (35,55–57). N-hydroxylation of the exocyclic amine group results in HONH-AαC, and ring oxidation of AαC at the C-3 and C-6 positions results in AαC-3-OH and AαC-6-OH, respectively (35,56). Various factors such as sex, race/Hispanic origin, and smoking status can affect CYP1A2 activity (58–60). Unfortunately, standards for these compounds and their corresponding stable isotope-labeled internal standards are not commercially available, so hydroxylated metabolites of AαC were not included in our assay. Therefore, the influence of race/Hispanic origin on AαC metabolism remains unclear and awaits future investigation.

We also found that certain diet components are significantly associated with AαC exposure. Beef and fish cooked at high temperature have been regarded as the most likely sources of AαC from foods eaten in the United States (34). Based on the formation mechanism, AαC is produced from pyrolysis of proteins or amino acids heated at high temperature (8). In the non-user and CPD regression models, eating beef cooked at high temperature was shown to be significantly associated with increased urinary AαC. In contrast, we found no significant association between urinary AαC and the amount of “meat, poultry, fish, and mixture” eaten without also specifying meat types or cooking temperature, which is consistent with a process requiring high temperature to contribute to dietary AαC exposure.

Eating vegetables was associated with significantly diminished urinary AαC levels in nonsmokers. Either decreased HAA absorption or enhanced hepatic elimination of HAA via induced CYP2A1 activity from eating vegetables has been proposed to explain reduced urinary HAA in persons who ate cooked meat (61,62). Because the hydroxylated AαC metabolites formed by CYP2A1 were not measured in our assay, we cannot say whether the influence of eating vegetables on urinary HAA levels in previous studies could be directly used to interpret our findings.

Fasting time was a significant negative predictor of urinary AαC levels in non-users of tobacco products, indicating that urinary excretion of AαC declined with time after the last meal. Because the AαC half-life is around 3.0 hr (63), urinary AαC levels would be expected to decline rapidly after complete AαC absorption from the food matrix.

We compared the potential influence of diet vs. smoking on urinary AαC by estimating the 99^th percentile of beef consumption cooked at high temperature for the United States population, which we found to be 0.22 kg/day. At this level of consumption, we estimated the likely increase in urinary AαC ΔY 95%CI to be 0.99 [0.16, 2.29]pg AαC/mL, compared to the increase associated with smoking one-half pack of cigarettes per day of 23.6 [18.7, 29.6] pg AαC/mL. Under these assumptions, the likely level of dietary exposure is considerably lower than from exclusive cigarette smoking.

Dietary exposure was based on 24-hour recall data collected in the MEC, thus reflecting dietary consumption for a single day. Although interviewers elicited recall with structured questions and standardized techniques, retrospective responses are generally susceptible to recall bias.

So far, no population-based study has identified tobacco smoke as a source of MeAαC exposure. A study in China lacked sufficient sensitivity of the urinary MeAαC assay to detect MeAαC in smokers (54). With the improved sensitivity of MeAαC measurement in our assay, our analysis provides the first population-based data on MeAαC. MeAαC was found to co-occur with AαC at about 10-fold lower content in tobacco smoke (7,8,10). As with AαC, MeAαC was mostly detected in smokers at around 20 times lower levels than those of AαC. Moreover, urinary AαC and MeAαC concentrations were significantly correlated among exclusive smokers with a sample-weighted Pearson correlation of 0.788. Therefore, MeAαC could serve as a valuable ancillary biomarker to assess AαC exposure.

In the United States, urinary AαC is significantly associated with serum cotinine levels among both exclusive smokers and non-users of tobacco, and tobacco smoke is a major source of AαC exposure in the general population. Much smaller increases in urinary AαC were associated with secondhand smoke exposure and eating beef cooked at high temperature.

ABBREVIATIONS

HAAs

Heterocyclic aromatic amines

AαC

2-amino-9H-pyrido[2,3-b]indole

MeAαC

2-amino-3-methyl-9H-pyrido[2,3-b]indole

NHANES

National Health and Nutrition Examination Survey

LC/MS/MS

liquid chromatography-tandem mass spectrometry

PIR

poverty-income ratio

CPD

cigarettes smoked per day

USDA

U.S. Department of Agriculture

BMI

body mass index