Evaluating the relationship between biomarkers of potential harm and biomarkers of tobacco exposure among current, past, and nonsmokers: data from the National Health and Nutrition Examination Survey 2007–2012

Abstract

Potential long-term health effects from tobacco products can be estimated by measuring changes in biochemical indicators of disease mechanisms like inflammation. This study assesses the potential relationships between biomarkers of potential harm (BOPH) and biomarkers of cigarette smoke exposure (BOE) based on data from the NHANES (2007–2012, n = 17,293 respondents). Statistically significant relationships were observed between white blood cells (WBC) and high-density lipoprotein (HDL) and BOE; between WBC and high-sensitivity C-reactive protein and smoking status; and between WBC and HDL and smoking intensity. This analysis suggests that WBC and HDL are useful BOPH in studies assessing the health risks of cigarette smoking.

Keywords:

• Biomarkers of exposure
• biomarkers of potential harm
• cigarette
• current smokers
• cigarette smoking
• National Health and Nutrition Examination Survey
• never smokers
• past smokers
• serum cotinine
• smoking status

Introduction

Cigarette smoking continues to be one of the greatest single preventable causes of premature mortality and morbidity. An estimated 480,000 annual deaths in the USA are related to tobacco use and exposure to smoke (American Lung Association, 2015). Cigarette smoking causes cardiovascular disease (CVD) (Gordon & Kannel, 1982), lung cancer, chronic obstructive pulmonary disease (COPD), emphysema, and other serious diseases (Centers for Disease Control and Prevention, 2010).

Evidence from the literature (Roethig et al., 2009; Sarkar et al., 2010) has demonstrated an association between cigarette smoking and biomarkers of inflammation. There is substantial evidence that there is a common mechanistic thread between the three major smoking-related diseases: lung cancer, COPD, and CVD. The 2010 Surgeon General’s Report (Centers for Disease Control and Prevention, 2010) states “Cigarette smoking produces a chronic inflammatory state that contributes to the atherogenic disease processes and elevates levels of biomarkers of inflammation, known powerful predictors of cardiovascular events.”

The increased risks of CVD are associated with elevated white blood cell (WBC) count and high sensitivity C-reactive protein (hs-CRP), which are considered markers of low-grade systemic inflammation (Madsen et al., 2007; Yasue et al., 2006). WBC count is a marker of inflammation and has been found to be an independent predictor of future coronary events (Abel et al., 2005; Brown et al., 2001). Several studies (Abel et al., 2005; Frost-Pineda et al., 2011; Liu et al., 2011) have shown an association between WBC count and smoking. However, the relationship between inflammatory markers and the biomarkers of cigarette smoke exposure (BOE) has not been systematically evaluated in large nationally representative samples.

Biomarkers of potential harm (BOPH) generally refer to a change in any level of the biological system due to exposure to a harmful substance (Frost-Pineda et al., 2011; Liu et al., 2011). The World Health Organization has recommended the use of the BOPH to assess potential long-term harmful effects that are otherwise difficult to observe in short-term controlled trials or prospective studies (World Health Organization, 2007). Some of the biomarkers that have shown a statistically significant relationship with smoking include WBC, hs-CRP, total cholesterol levels, high-density lipoprotein (HDL), and low-density (LDL) (Frost-Pineda et al., 2011). BOE can be either a chemical compound or its metabolite that reflects the internal dose of exposure to tobacco constituents (World Health Organization, 2007). Biomarkers of exposure (BOEs) in the urine or blood of adult cigarette smokers can provide quantitative estimates of the uptake of selective smoke constituents (Hatsukami et al., 2006; Institute of Medicine, 2001).

Serum cotinine, the primary metabolite of nicotine, is one of the most widely used BOE in tobacco exposure studies (Benowitz & Jacob, 1994; O'Connor et al., 2006). Several tobacco-specific nitrosamines are found in tobacco and cigarette smoke, and of these, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) has been classified as a Group 1 carcinogen (International Agency for Research on Cancer, 2010). NNK exposure can be measured by its metabolites 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL) and NNAL glucuronide conjugates, which are excreted in urine (Hecht, 2002; Hecht et al., 1993). There are hundreds of polyaromatic hydrocarbons (PAHs) and their isomers in the particulate phase of cigarette smoke (Pelkonen & Nebert, 1982; Rodgman et al., 2000). Pyrene, although not carcinogenic, is one of the PAHs found in cigarette smoke and is often used as an indicator of total PAH exposure (Brandt & Watson, 2003; Jongeneelen, 1994). It is primarily (∼90%) metabolized to 1-hydroxypyrene (1-OHP) by CYP1A enzymes, and it is eliminated in the urine as glucuronide and sulfate conjugates (total 1-OHP). Although this biomarker is ubiquitous and not specific to cigarette smoke (Hatsukami et al., 2006; Johnson et al., 2010b), it has been used extensively for bio-monitoring of PAHs, and exposure from sources other than cigarette smoke is well established (Roethig et al., 2009). Furthermore, PAHs have been suggested to be involved in smoking-related diseases (Yuan et al., 2014), therefore we wanted to investigate the relationship between this biomarker and BOPHs.

Currently in the USA, national and state prevalence data on cigarette smoking for adolescents and adults are gathered using questionnaires that rely on self-reported information to assess smoking status. The National Health and Nutrition Examination Survey (NHANES) is one such nationally administered survey designed to determine the health and nutritional status of US adults and children (Centers for Disease Control, 2015). The survey is conducted by the National Center for Health Statistics, which is a part of Centers for Disease Control and Prevention, and it examines a nationally representative sample of about 5000 persons each year. The NHANES interview includes variables on demographic, socioeconomic, dietary, lifestyle, and health-related aspects of the participants. The survey is unique in that it combines interviews with physical examinations and biochemical assessment of body fluids, including blood serum, and data on several BOPH and BOE. The sampling design is based on a complex, multistage probability strategy that includes the selection of primary sampling units (counties), household segments within the counties, and sample participants from selected households (Centers for Disease Control, 2015).

The primary objective of this study was to examine the quantitative relationships between the levels of six BOPH (WBC, apolipoprotein, HDL, LDL, hs-CRP, and total cholesterol) and the three BOE (serum cotinine, urinary NNAL, and 1-OHP) among current, past, and never smokers based on the NHANES data over a 6-year period from 2007 to 2012. The secondary objective was to assess the relationship between BOPH and smoking status (current, past, or never) of the respondent based on self-reported responses to the NHANES questionnaire. BOPH in current smokers were also tested for the association with cigarettes per day (CPD).

Methods

Data were obtained from the NHANES, which examines a nationally representative sample of US non-institutionalized population. The survey examines about 5000 persons each year and uses a complex, stratified, multistage, probability cluster design that oversamples low-income persons, adolescents 12–19 years of age, persons above 60 years of age, African Americans, and Mexican Americans. The NHANES data from years 2007–2008, 2009–2010, and 2011–2012 were merged by unique identification numbers to construct a comprehensive 6-year dataset. The sample population comprised 30,344 subjects who answered questions on tobacco usage, completed the modules for the smoking questionnaire, and had recorded laboratory values for the selected biomarkers collected by the mobile examination center (MEC). Demographic data including age, gender, race/ethnicity, and body mass index (BMI) were derived from self-reported individual and household questionnaire data. For the purpose of this study, we categorized the sample population in age groups of 21–35, 36–50, 51–65, and >65 years. We limited our comparisons with observations from respondents of age 21 and above since majority of the participants under age 21 belonged to the age category of <8 years (53.65%) with 10% being categorized as <1 year old. We wanted to minimize any potential confounding effects due to differences in xenobiotic metabolism within the lower age groups (Ginsberg et al., 2004; Graeter & Mortensen, 1996). Furthermore, we have conducted analysis for participants ≥18 years of age to select biomarkers of exposure and biomarkers of potential harm (see Supplemental Table 1S, 2S and 3S). Indeed, our analysis confirms that there is very little difference between the biomarker levels as well as the regression analysis for age ≥18 years and ≥21 years of age. We did not apply any additional exclusion criteria other than limiting the dataset to age <21 years to get to the final sample size. Missing data for any variable were treated as missing observations and excluded from the regression analyses through the SAS procedure.

The race/ethnicity variable also was categorized into four groups, including Hispanic, black non-Hispanic (NH), white NH, and other NH participants. To examine the effect of smoking status on the levels of BOPH, the sample population was categorized into current, past, and never smokers. Respondents who reported having smoked at least 100 cigarettes in their entire life and were currently smoking every day or some days at the time of interview were classified as current smokers (CS). Respondents who reported smoking at least 100 cigarettes during their lifetime but currently did not smoke for the past 1 year were classified as past smokers (PS). Although the half-lives of some of the BOEs are relatively short (e.g. cotinine half-life is ∼18–20 h: Benowitz & Jacob, 2001), NNAL has a significantly long half-life (∼45 d: Hecht et al., 1999). However, the main rationale for the past 1-year criterion was that we also wanted to allow for a long-enough time to elapse to avoid the residual impact of previous smoking on the BOPH. We have reported that the BOPHs investigated in this analysis are significantly reduced in 4 weeks and stabilize in about 3–12 months (Roethig et al., 2008). Respondents who reported not having smoked at least 100 cigarettes in their life and were not currently smoking were classified as never smokers (NS). This method of smokers’ classification has been recommended by NHANES and has been used in previous studies investigating smoking populations (Albanes et al., 1987; Clair et al., 2011).

The analytical limit of detection of the NHANES method for measuring serum cotinine is 0.015 ng/mL, urinary NNAL as 0.6 pg/mL and for 1-hydroxypyrene as 7.1 ng/L. Benowitz et al. (2009) reported that the optimal serum cotinine level for distinguishing cigarette smokers and nonsmokers is 3 ng/mL. We determined the percentages of PS and NS that were above the NHANES analytical limits of detection of the three BOEs. We also estimated the proportion of respondents who were above the threshold of the optimal serum cotinine level of 3 ng/mL. The serum cotinine cutoff used in this analysis (3 ng/mL) may not reflect the optimal levels across the race/ethnicities in this study.

To minimize the variability in the concentrations of urinary biomarkers, due to the differences in urine volume, urine concentrations were normalized against creatinine. Levels of the urinary BOE, NNAL and 1-OHP, were normalized by dividing the measured level by the corresponding urine creatinine concentration (Sarkar et al., 2008). The BOPH variables were analyzed as continuous variables. To assess the effect of CPD use on all the BOPH, a five-level categorical variable (1–10; 11–20; 21–30; 30–40; and >40) based on the number of cigarettes smoked per day in CS was constructed. The number of CPD was the respondent self-reported average number of cigarettes smoked in the past 30 d.

Statistical analysis

Descriptive analysis was performed on age, gender, race/ethnicity, smoking status, BMI, and CPD (only for CS) for the sample populations. Sampling weights were derived from the unequal probability of selection and non-response adjustment to make the sample nationally representative (NHANES, 2015). Both un-weighted numbers and weighted percentages were reported in the descriptive analysis. Univariate analyses were performed by calculating the mean concentrations and range of the BOE and BOPH among CS, PS, and NS. Since all the model response variables were in continuous format, multivariate survey linear regression models were used to assess the relationship between the response variable and exposure variables, adjusting for confounding effects of demographic covariates. Missing data for any variable were treated as missing observations and excluded from the regression analyses. The regression models accounted for the NHANES cluster sampling technique by using sampling clusters, strata, and weights in the models.

For the first set of analyses, the independent variables were the BOE’s serum cotinine, NNAL (creatinine adjusted), and 1-OHP (creatinine adjusted); and the outcome variables were the selected BOPH. For the second set of analyses, the independent variables were smoking status of respondents and CPD use, while the outcome variables were the BOPH. All statistical analyses were performed using SAS (version 9.4, SAS Institute Inc., Cary, NC).

Results

After excluding smokers under 21 years of age, the analytical sample had 17,293 respondents. Of that sample, 20.93% were CS, 24.38% were PS, and 54.69% were NS. Table 1 shows the demographic distribution of the nationally representative sample. The mean BMI of the sample population was 29.02 kg/m², and around 54% of CS smoked between 1 and 10 CPD. Tables 2 and 3 show the distribution of the BOE and BOPH, respectively, among CS, PS, and NS. As expected, serum cotinine (218.37 ng/mL), urinary NNAL (383.73 ng/mg Cr) and 1-OHP (341.34 ng/mg Cr) concentrations were found to be higher in CS than in PS and NS. Serum cotinine levels were above the limit of detection (0.015 ng/mL) in 65.72% NS and 72.39% PS. However, the majority of PS (89.02%) and NS (94.91%) had serum cotinine levels below the threshold of the optimal serum cotinine level of 3 ng/mL used to differentiate CS and NS. The distribution of NS and PS were 46.4% and 52.92%, respectively, for urinary NNAL (limit of detection of 0.6 pg/mL) and were 98.83% and 98.46%, respectively, for 1-OHP (limit of detection of 7.1 ng/mL).

Table 1. Descriptive statistics of demographics in the study population.

Population characteristics	N^a	Weighted %^b
Age
21–35 years	4368	27.82
36–50 years	4503	30.13
51–65 years	4386	25.76
>65 years	4036	16.29
Race/ethnicity
Hispanic/Latino	4494	13.72
Black NH	3689	11.28
White NH	7604	68.01
Other NH	1506	6.99
Gender
Male	8432	47.86
Female	8861	52.14
Smoking status
Past smoker	4215	24.38
Current smoker	3663	20.93
Never smoker	9399	54.69
Cigarettes per day (only for current smokers)
1–10	2139	54.03
11–20	1112	33.07
21–30	229	8.10
30–40	117	3.83
>41	42	0.97
Body mass index (kg/m^2)	Mean (SD)	Range
	29.02 (6.78)	13.18–84.87

NH: non-Hispanic; SD: standard deviation.

a N: total number of participants in each category without using the sampling weights.

b Weighted %: percentage of participants in each category after attaching the sampling weights.

Table 2. Descriptive statistics for biomarkers of exposure (BOE) by smoking status.

BOE	Smoking status	N	Mean	Standard error	Minimum	Maximum
Serum cotinine (ng/mL)	NS	8439	6.75	0.77	0.01	599.00
	CS	3300	218.37	4.56	0.01	597.00
	PS	3841	19.66	1.96	0.01	599.00
Adjusted NNAL (ng/mg Cr)	NS	8245	27.18	3.53	0.08	33,762.00
	CS	3312	383.73	17.36	0.10	4138.88
	PS	3653	66.49	11.63	0.09	43,243.00
Adjusted PAH (ng/mg Cr)	NS	1871	134.91	7.60	2.92	3636.20
	CS	696	341.34	15.17	5.07	3102.04
	PS	833	143.16	11.17	3.30	5666.66

CS: current smokers; NNAL: 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol; NS: never smokers; PAH: polyaromatic hydrocarbons; PS: past smokers.

Table 3. Descriptive statistics for biomarkers of potential harm (BOPH) by smoking status.

BOPH	Smoking status	N	Mean	Standard error	Minimum	Maximum
WBC count (×1000 cells/μL)	NS	8516	6.82	0.037	1.6	43.9
	CS	3350	8.15	0.072	1.5	37.4
	PS	3878	6.98	0.052	1.7	42.5
Direct HDL-cholesterol (mg/dL)	NS	8417	53.54	0.266	7	156
	CS	3290	49.93	0.481	15	175
	PS	3835	53.23	0.408	13	179
LDL-cholesterol (mg/dL)	NS	4118	116.12	0.688	9	344
	CS	1524	117.11	1.261	24	341
	PS	1833	114.70	0.883	15	290
Total cholesterol (mg/dL)	NS	8417	196.30	0.626	78	475
	CS	3290	197.16	1.012	59	528
	PS	3835	197.77	0.959	75	413
Apolipoprotein (B) (mg/dL)	NS	4188	90.66	0.626	24	215
	CS	1568	94.04	0.912	28	260
	PS	1875	92.02	0.654	15	253
hs-CRP (mg/dL)	NS	5717	0.35	0.01	0.01	13.9
	CS	2363	0.45	0.022	0.01	12.2
	PS	2725	0.41	0.011	0.01	13.1

WBC: white blood cell; NS: never smokers; CS: current smokers; PS: past smokers; HDL: high-density lipoproteins; LDL: low-density lipoproteins; hs-CRP: highly sensitive C-reactive protein.

BOPH distribution showed that CS had a higher level of WBC (8.15 × 1000/μL), LDL (117.11 mg/dL), apolipoprotein (94.04 mg/dL) and hs-CRP, (0.45 mg/dL), and lower HDL (49.93 mg/dL) concentrations than PS and NS.

Table 4 reports the regression result between serum cotinine and the BOPH. After controlling for age, sex, race/ethnicity, and BMI, it was observed that the levels of WBC (p < 0.001, r² = 0.133), HDL (p < 0.001, r² = 0.222), and hs-CRP (p = 0.033, r² = 0.107) showed a statistically significant relationship with serum cotinine. The relative ranking of the serum cotinine, based on the F-value from the regression model for the BOPH, was WBC (178.74) > HDL (35.88) > hs-CRP (4.95). Serum cotinine was the most important factor in the model for WBC. Among other covariates, age, race/ethnicity, and BMI showed a significant relationship with WBC, of which BMI had the most significant impact in the model (F-value =74.07) after serum cotinine. The model for HDL included age, race/ethnicity, sex, and BMI, with sex having the most significant impact (F-value = 224.51). Sex and BMI were the statistically significant covariates for hs-CRP, with BMI having the most significant impact (F-value = 226.49).

Table 4. Regression model results with serum cotinine.

	WBC		LDL		HDL		Total cholesterol		Apolipoprotein		hs-CRP
	F-value		F-value		F-value		F-value		F-value		F-value
Parameter	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value
Serum cotinine	178.74	<0.0001	0.53	0.4715	35.88	<0.0001	0.50	0.4837	0.70	0.4093	4.95	0.0332
	(0.005)		(0.005)		(−0.016)		(−0.006)		(0.004)		(0.0005)
Age	9.31	0.0046	10.89	0.0024	10.56	0.0027	27.40	<0.0001	29.63	<0.0001	0.24	0.6264
	(−0.160)		(3.577)		(1.004)		(6.674)		(3.780)		(0.008)
Race/ethnicity	15.68	0.0004	0.07	0.7870	14.19	0.0007	1.23	0.2762	3.11	0.0875	1.95	0.1720
	(−0.232)		(−0.318)		(1.437)		(−1.427)		(−1.521)		(−0.022)
Sex	0.34	0.5653	2.16	0.1511	224.51	<0.0001	3.54	0.0691	6.20	0.0182	10.14	0.0032
	(0.052)		(−2.629)		(9.672)		(4.176)		(−3.504)		(0.091)
BMI	74.07	<0.0001	4.86	0.0348	172.65	<0.0001	0.77	0.3878	28.81	<0.0001	226.49	<0.0001
	(0.075)		(0.276)		(−0.772)		(−0.114)		(0.501)		(0.037)
R^2	0.133		0.015		0.222		0.032		0.053		0.107

WBC: white blood cells; HDL: high-density lipoproteins; LDL: low-density lipoproteins; BMI: body mass index; hs-CRP: highly sensitive C-reactive protein.

p < 0.05 was considered statistically significant, a priori. F-value refers to the relative impact of each variable within the individual model. β estimate refers to the parameter estimates of the regression model. R² refers to the overall model fit.

Table 5 lists the regression results between adjusted urinary NNAL and the BOPH. There was a significant association between NNAL and WBC (p = 0.0489, r² = 0.072) and HDL (p = 0.017, r² = 0.210). Age, race/ethnicity, and BMI were the other covariates that were statistically significantly associated with WBC, with BMI having the greatest impact (F-value = 48.51). Whereas age, race/ethnicity, sex, and BMI were significantly associated with HDL and sex having the highest weight (F-value = 225.83).

Table 5. Regression model results with adjusted NNAL.

	WBC		LDL		HDL		Total cholesterol		Apolipoprotein		hs-CRP
	F-value		F-value		F-value		F-value		F-value		F-value
Parameter	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value
Adjusted NNAL	4.19	0.0489	0.02	0.8789	6.34	0.0170	0.0001	0.9903	0.30	0.5908	1.69	0.203
	(0.001)		(0.0004)		(−0.002)		(0.0003)		(0.001)		(0.0001)
Age	13.59	0.0008	10.62	0.0026	14.55	0.0006	26.97	<0.0001	29.10	<0.0001	0.04	0.840
	(−0.202)		(3.536)		(1.143)		(6.713)		(3.763)		(0.003)
Race/ethnicity	7.38	0.0106	0.04	0.8387	9.70	0.0039	1.41	0.2435	2.96	0.0952	0.84	0.366
	(−0.161)		(−0.238)		(1.200)		(−1.519)		(−1.462)		(−0.014)
Sex	0.05	0.8316	2.24	0.1445	225.83	<0.0001	3.78	0.0608	6.32	0.0172	6.60	0.015
	(−0.022)		(−2.725)		(9.923)		(4.299)		(−3.562)		(0.082)
BMI	48.51	<0.0001	4.57	0.0402	146.25	<0.0001	0.60	0.4437	29.19	<0.0001	237.74	<0.0001
	(0.065)		(0.264)		(−0.737)		(−0.098)		(0.493)		(0.036)
R^2	0.072		0.015		0.210		0.032		0.053		0.101

WBC: white blood cells; HDL: high-density lipoproteins; LDL: low-density lipoproteins; BMI: body mass index; hs-CRP: highly sensitive C-reactive protein; NNAL: 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol.

p < 0.05 was considered statistically significant, a priori. F-value refers to the relative impact of each variable within the individual model. β estimate refers to the parameter estimates of the regression model. R² refers to the overall model fit.

Table 6 presents the regression results between 1-OHP (adjusted) and the BOPH. It was observed that WBC (p = 0.041, r² = 0.090) and HDL (p = 0.0059, r² = 0.186) showed a significant relationship with 1-OHP, after controlling for other covariates. Race/ethnicity and BMI were the other covariates that were significantly associated with WBC, of which BMI had the highest impact (F-value = 7.90). Sex and BMI were significantly associated with HDL, with sex being the most significant factor (F-value = 41.81).

Table 6. Regression model results with adjusted 1-OHP.

	WBC		LDL		HDL		Total cholesterol		Apolipoprotein		hs-CRP
	F-value		F-value		F-value		F-value		F-value		F-value
Parameter	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value
Adjusted 1-OHP	4.55	0.0406	0.03	0.8585	8.70	0.0059	0.49	0.487	0.03	0.8749	0.88	0.3565
	(0.003)		(0.003)		(−0.013)		(−0.014)		(−0.002)		(0.0002)
Age	0.0001	0.9961	2.22	0.1462	0.18	0.6708	4.60	0.040	6.92	0.0130	2.58	0.1177
	(0.0003)		(2.731)		(0.339)		(4.868)		(3.868)		(0.047)
Race/ethnicity	5.41	0.0266	0.09	0.7618	2.36	0.1345	0.26	0.613	1.00	0.3238	0.80	0.3782
	(−0.290)		(−0.972)		(1.881)		(−1.509)		(−1.829)		(−0.017)
Sex	0.33	0.5693	0.54	0.4663	41.81	<0.0001	0.29	0.594	2.17	0.1501	3.59	0.0672
	(0.149)		(−3.109)		(8.182)		(1.974)		(−3.756)		(0.157)
BMI	7.90	0.0083	0.08	0.785	11.36	0.0020	0.05	0.826	1.62	0.2118	18.10	0.0002
	(0.076)		(0.124)		(−0.475)		(−0.107)		(0.355)		(0.030
R^2	0.090		0.011		0.186		0.023		0.056		0.190

1-OHP: 1-hydroxypyrene; WBC: white blood cells; HDL: high-density lipoproteins; LDL: low-density lipoproteins; BMI: body mass index; hs-CRP: highly sensitive C-reactive protein.

p < 0.05 was considered statistically significant, a priori. F-value refers to the relative impact of each variable within the individual model. β estimate refers to the parameter estimates of the regression model. R² refers to the overall model fit.

Table 7 presents the regression results between BOPH and smoking status. After controlling for age, sex, race/ethnicity, and BMI, it was observed that smoking status had significant relationships with WBC (p = 0.0002, r² = 0.060) and hs-CRP (p = 0.0399, r² = 0.101). For the WBC model, age, race/ethnicity, and BMI also showed a significant relationship, with BMI showing the highest impact (F-value = 45.90). Sex and BMI showed a significant relationship for hs-CRP with BMI having the highest impact (F-value = 237.41) in the model.

Table 7. BOPH regression model estimates with smoking status.

	WBC		LDL		HDL		Total cholesterol		Apolipoprotein		hs-CRP
	F-value		F-value		F-value		F-value		F-value	p-value	F-value
Parameter	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value
Smoking status	17.40	0.0002	0.97	0.3320	0.06	0.8033	0.28	0.6015	0.15	0.7035	4.59	0.0399
	(−0.269)		(1.281)		(0.098)		(0.658)		(0.279)		(−0.039)
Age	22.04	<0.0001	12.04	0.0015	14.40	0.0006	29.50	<0.0001	31.19	<0.0001	0.02	0.8909
	(−0.238)		(3.750)		(1.134)		(6.821)		(3.816)		(−0.002)
Race/ethnicity	7.42	0.0104	0.03	0.8560	9.26	0.0046	1.39	0.2477	2.91	0.0976	0.82	0.3729
	(−0.157)		(−0.211)		(1.174)		(−1.508)		(−1.448)		(−0.014)
Sex	0.02	0.9006	2.51	0.1228	207.86	<0.0001	3.42	0.0736	6.21	0.0181	6.82	0.0136
	(−0.014)		(−2.935)		(10.010)		(4.199)		(−3.636)		(0.085)
BMI	45.90	<0.0001	4.58	0.0402	138.11	<0.0001	0.69	0.4129	30.25	<0.0001	237.41	<0.0001
	(0.063)		(0.254)		(−0.728)		(−0.103)		(0.488)		(0.036)
R^2	0.060		0.015		0.206		0.032		0.053		0.101

WBC: white blood cells; HDL: high-density lipoproteins; LDL: low-density lipoproteins; BMI: body mass index; hs-CRP: highly sensitive C-reactive protein.

p < 0.05 was considered statistically significant, a priori. F-value refers to the relative impact of each variable within the individual model. β estimate refers to the parameter estimates of the regression model. R² refers to the overall model fit.

Table 8 lists the regression model with CPD as the predictor variable wherein WBC (p = 0.0015, r² = 0.098) and HDL (p = 0.0005, r² = 0.160) showed a significant relationship. BMI was the only other statistically significant covariate for WBC. Both sex and BMI were significant factors in the HDL model, with BMI having the most significant impact (F-value = 87.13).

Table 8. BOPH regression model estimates with cigarette per day.

	WBC		LDL		HDL		Total cholesterol		Apolipoprotein		hs-CRP
	F-value		F-value		F-value		F-value		F-value		F-value
Parameter	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value	(β estimate)	p-value
CPD	12.17	0.0015	0.01	0.9295	15.00	0.0005	0.14	0.7068	0.19	0.6682	1.64	0.2101
	(0.663)		(0.261)		(−3.274)		(−1.204)		(0.836)		(0.114)
Age	0.02	0.8787	1.78	0.1920	1.030	0.3175	4.69	0.0381	4.24	0.0479	0.86	0.3610
	(−0.016)		(3.724)		(1.019)		(6.807)		(4.002)		(0.052)
Race/ethnicity	0.77	0.3870	0.01	0.9354	1.54	0.2237	0.20	0.6566	0.22	0.6387	0.02	0.8784
	(−0.149)		(0.219)		(1.158)		(−1.419)		(−0.815)		(0.008)
Sex	0.16	0.6934	0.45	0.5082	15.23	0.0005	0.03	0.8599	0.90	0.3504	1.09	0.3046
	(0.096)		(−2.636)		(5.639)		(1.001)		(−3.185)		(−0.119)
BMI	16.05	0.0004	1.58	0.2187	87.13	<0.0001	0.05	0.8206	14.65	0.0006	22.88	<0.0001
	(0.085)		(0.292)		(−0.833)		(−0.063)		(0.640)		(0.040)
R^2	0.098		0.013		0.160		0.021		0.055		0.075

WBC: white blood cells; HDL: high-density lipoproteins; LDL: low-density lipoproteins; CPD: cigarettes per day use; BMI: body mass index; hs-CRP: highly sensitive C-reactive protein.

p < 0.05 was considered statistically significant, a priori. F-value refers to the relative impact of each variable within the individual model. β estimate refers to the parameter estimates of the regression model. R² refers to the overall model fit.

Discussion

Our study assessed the relationship of BOPH with BOE, smoking status, and CPD use among, current, past, and never smokers from a non-institutionalized adult population in the USA during 2007–2012. After adjusting for covariates (age, race/ethnicity, sex, and BMI), WBC demonstrated a consistent association with serum cotinine, urinary total NNAL, and 1-OHP, as well as smoking status and CPD. Similarly a consistent relationship was observed with all the BOE and CPD use for HDL, whereas hs-CRP demonstrated associations only with serum cotinine and smoking status.

As expected, BOE levels were found to be highest in CS than PS and NS. It was also not surprising that detectable levels of the 1-OHP were observed in PS and NS, since exposure to PAHs from sources other than cigarette smoke is well established (Strickland & Kang, 1999). Researchers have reported significant levels of 1-OHP in nonsmokers (St. Helen et al., 2012). However, it was surprising that detectable levels of the tobacco-specific biomarkers (serum cotinine and total NNAL) were observed in a relatively large proportion of PS and NS. Similar observations have been reported by other investigators (Benowitz et al., 2009; Roethig et al., 2009). This could potentially be attributed to the misclassification of respondents as NS and particularly PS (Jedrychowski et al., 1998). For example, additional analysis of the data revealed that 4.47% of PS and 2.28% of NS had used either pipe, cigars, chewing tobacco, snuff, nicotine patch, or nicotine gum in the last 5 d. Exposure to secondhand smoke may account for the measurable levels observed in NS and PS. In a recent report (Jain, 2015) based on the analysis of the 2011–2012 NHANES data of nonsmokers, exposure to secondhand smoke at home was associated with serum cotinine levels about 30 times higher than those without such exposure (0.717 ng/mL vs. 0.024 ng/mL, p < 0.01). NNAL levels among nonsmokers being exposed to secondhand smoke at home were about 20 times higher than the levels among nonsmokers without such exposure (9 pg/mL vs. 109 pg/mL, p < 0.01). However, as the magnitude of secondhand smoke exposure declines because of proportionally fewer smokers and more clean-indoor-air regulations, the optimal cotinine threshold for distinguishing smokers from nonsmokers has changed. In a study analyzing NHANES data from 1999 to 2004, Benowitz et al. (2009) suggest that the threshold of the optimal serum cotinine level should be 3 ng/mL. Our analysis suggests that this might, indeed, be more appropriate since a large majority of the PS and NS had levels <3 ng/mL (89% and 95%, respectively). It is possible that the respondents with serum cotinine levels >3 ng/mL were either misclassified or could be using other tobacco or nicotine-containing products.

Our results demonstrate that after adjusting for age, sex, race/ethnicity, and BMI, a statistically significant relationship was observed between levels of WBC, HDL, and hs-CRP and serum cotinine levels. These results are consistent with other published reports. A multi-center study conducted in a population of adult smokers and nonsmokers by Liu et al. (2011) found a similar statistically significant relationship between cigarette smoke exposure and WBC count.

The direction of the relationships between BOE and BOPH reported here (i.e. higher WBC and lower HDL levels as BOE levels increases) is supported by the biological plausibility of these end-points. Our results demonstrate that CPD use was positively associated with WBC levels and negatively associated with HDL levels. These results corroborate the findings of other studies (Frost-Pineda et al., 2011; Liu et al., 2011), which showed a similar relationship between WBC and HDL levels with CPD.

Our observation of serum cotinine being the most significant factor in the regression model with WBC is consistent with other reports. For example, Liu et al. (2011) reported that urinary nicotine equivalents (nicotine and five of its metabolites) were the most important factor in a regression model with WBC. These results establish that WBC might be an important biomarker when investigating the impact of changes with cigarette smoke exposure. Furthermore, there is a substantial bioplausible explanation of the role of WBC in both CVD and pulmonary disease. An increase in WBC count is one of the most important early events in the initiation of the disease (Lee et al., 2009). Adhesion of circulating WBC to the endothelium is among the first steps in the initiation of the atherosclerosis, followed by directed migration of the bound WBC into the intima, maturation of the WBC into macrophages, and their uptake of lipid, yielding foam cells (Bevilacqua et al., 1985). The progression of events to thrombosis, thereafter, is also well established (Bevilacqua et al., 1985). Furthermore, WBC are not only involved in progression of CVD, but also play an important role in pulmonary injury as well. Chronic inflammation has been linked to COPD, emphysema, and lung cancer (Ind, 2005; Lee et al., 2009). Alveolar macrophages are elevated in the lungs of smokers, suggesting a role in combatting pulmonary inflammation (Martinez-Sanchez et al., 2014).

Not only is WBC count biologically relevant as a BOPH, but this biomarker is also reversible upon smoking cessation, and levels are reduced when CS switch from smoking cigarettes to a reduced-exposure smoking system (Roethig et al., 2008). Statistically significant and clinically favorable changes in WBC were observed within the first 4 weeks in a randomized controlled clinical trial in which adult smokers were switched to an electrically heated cigarette system (Roethig et al., 2008). Similar changes in WBC count were reported in a population of smokers who reduced their smoking of cigarettes by 50% (Eliasson et al., 2001). Therefore, we believe that changes in the levels of WBC are an important indicator of changes in individual health risk, particularly since WBC has been reported to be an independent predictor of death from CVD. A decrease in WBC count of 1000 cells/μL is reportedly associated with a decrease of 14% in the risk of death from CVD (Brown et al., 2001; Margolis et al., 2005; Tamakoshi et al., 2007).

We observed a significant relationship between HDL and BOE (serum cotinine and total NNAL), which is consistent with other published reports (Liu et al., 2011). However, some studies (Calapai et al., 2009; Lowe et al., 2009) report lack of a significant change between smokers and nonsmokers, which could possibly be due to the relatively small sample size (n = 20) per group. The relationship between HDL and BOE is statistically significant in studies with larger sample sizes, e.g. n = 17,293 (in our study) and n = 3585 smokers and n = 1077 nonsmokers (Liu et al., 2011). The results from our model suggest that HDL is not specific to cigarette smoking and other factors influence this biomarker. Nevertheless, in a recent review of 45 studies assessing HDL levels upon smoking cessation, Forey et al. (2013) concluded that “Quitting smoking is clearly associated with an increase in HDL concentrations. Generally the increase occurs rapidly, in less than three weeks, with no clear pattern of change thereafter.” Similar changes have been observed in a randomized controlled clinical trial in which adult smokers were switched to an electrically heated cigarette system (Roethig et al., 2008). Statistically significant and clinically favorable changes in HDL were observed within 4 weeks, without any subsequent major changes for the rest of the study duration for up to 12 months (Roethig et al., 2008).

It is well established that HDL is a significant factor that inhibits atherosclerosis at several key stages (Hessler et al., 1979). HDL inhibits the oxidation of LDL, which is one of the first steps in the progression of atherosclerosis (Hessler et al., 1979). It has also been reported that HDL promotes cholesterol efflux from macrophages in the arterial wall, thereby limiting the inflammatory process that underlies atherosclerosis (Hessler et al., 1979; Tall et al., 2008). HDL has several additional protective properties that are independent of cholesterol metabolism, such as reducing oxidation, vascular inflammation and thrombosis; improving endothelial function; promoting endothelial repair; enhancing insulin sensitivity; and promoting insulin secretion by pancreatic islet beta cells (Barter, 2011). There is also a large and compelling body of evidence in animal models showing that interventions that increase HDL levels are profoundly anti-atherogenic (Forte et al., 2002; Navab et al., 2011). It has also been suggested that HDL is an independent risk predictor of CVD. Epidemiological studies suggest that for every for every 2–3% increase in the level of HDL (independent of LDL), there is a 2–4% reduction in CVD events (Charland & Malone, 2010). Therefore, we believe that a change in HDL level will be an important indicator of changes in individual health risks in smokers.

Since both HDL and WBC are well-established and well-accepted biomarkers as potentially predictive of disease risk, they will play important roles in future clinical studies assessing the health risks of cigarette smoking. It is important to note that while statistically significant and favorable changes were observed with smoking measures, other factors had a greater impact on the BOE. For example, BMI clearly had a significant impact on most of the BOPHs. Similar observations have been reported in a large-scale cross-sectional study (Frost-Pineda et al., 2011). This is not surprising since obesity, as indicated by a high BMI, is a significant risk factor for several diseases including type 2 diabetes, coronary heart disease, cancers (endometrial, breast, and colon), hypertension, and stroke (National Institutes of Health, 1998).

Our results do not show a statistically significant relationship between BOE, smoking status, and CPD with LDL. This is supported by similar reports of insufficient evidence of a relationship between LDL and cigarette smoke exposure. The differences between smokers and nonsmokers has been reportedly small and only borderline statistically significant (Brandt & Watson, 2003; Frost-Pineda et al., 2011). In a study of adult smokers (n = 1504), LDL levels were not significantly altered in a subgroup of abstainers (n = 334) (Johnson et al., 2010a). Given the lack of consistent and insufficient evidence of relationship between cigarette smoke exposure and LDL levels, we suggest that LDL may not be considered a suitable biomarker for future studies in smokers.

Similarly, the lack of a consistent relationship with hs-CRP confirms observations from a large study of smokers (n = 1504), in which smoking intensity was not associated with hs-CRP levels and smoking cessation did not reduce hs-CRP levels (Johnson et al., 2010b). In a review (Hlatky et al., 2009) on the assessment of hs-CRP as a biomarker to predict coronary atherosclerosis, based on the criteria developed by the American Heart Association, the authors concluded that “advocating for the widespread use of hs-CRP for CV [cardiovascular] prediction is premature” (Hlatky et al., 2009). Therefore, it appears that hs-CRP would not be a suitable biomarker to be included in studies assessing health risks in smokers. Lack of associations with any of the smoking measures also suggests that apo-lipoprotein is not discriminatory and therefore would not be a suitable biomarker for smoking studies.

The results of this study should be interpreted in consideration of some of the limitations. First, since the estimates of cigarette smoking were self-reported, there could be a potential of recall bias. Secondly, the data used in the analysis were from cross-sectional surveys and do not represent longitudinal follow-up. Thirdly, we did not examine the relationship between smoking patterns (either daily/some days) and cardiovascular biomarker levels among our population. The underlying assumption for not examining this relationship was based on reports that BOEs are related to number of cigarettes smoked per day but not number of days per month of smoking (Khariwala et al., 2014). Therefore, it is unlikely that the relationship between BOE and BOPH would be very different. Nevertheless, this could be a topic for future analysis. Lastly, the strength of the relationship is weak (r² ranging from 0.02 to 0.13), suggesting that only a small fraction of the variability is explained by the factors included in the model. Nevertheless, some of the relationships are relatively stronger than others and are bioplausible as well as consistent with other reports in the literature.

Future analyses could examine the relationships between smoking behavior (e.g. different types of cigarettes, duration of smoking) and cardiovascular biomarker levels in current smokers. Also, the influence of polytobacco use (e.g. cigars, smokeless tobacco, etc.) could be included to examine the variability in the relationship. Despite these limitations, our use of a nationally representative sample in this study strengthens our understanding of the relationship between BOE and BOPH. The consistency in the results reported here, based on a nationally representative survey, supports the evidence that WBC and HDL are important biomarkers to be included in future studies investigating changes in health risks from changes in exposure in cigarette smokers.

Conclusion

Identifying and studying BOPH is important for assessing the potential long-term harmful effects of cigarette smoking. The primary objective of this study was to assess the potential relationships between BOPH (specifically WBC, apolipoprotein, hs-CRP, HDL, LDL, total cholesterol) and BOE (specifically serum cotinine, creatinine adjusted urinary total NNAL and 1-OHP) using the NHANES data from 2007 to 2012. The secondary objective was to assess the relationship between BOPH and smoking status (current, past, or never), and CPD use in CS. Among all the BOPH, the correlation between WBC and HDL were significant with all the cigarette-smoking measures. These observations confirm the utility of WBC and HDL in studies assessing the health risks of cigarette smoking. The results of our study were consistent with published studies that used primary data, indicating that national level survey datasets such as the NHANES may be a useful source for identifying biomarkers for use in studies of tobacco product users.

Abbreviations
1-OHP	=	1-hydroxypyrene
BMI	=	body mass index
BOE	=	biomarkers of cigarette smoke exposure
BOPH	=	biomarkers of potential harm
COPD	=	chronic obstructive pulmonary disease
CPD	=	cigarettes per day
CS	=	current smokers
CVD	=	cardiovascular diseases
HDL	=	high-density lipoproteins
hs-CRP	=	high-sensitivity C-reactive protein
LDL	=	low-density lipoproteins
NH	=	non-Hispanic
NHANES	=	National Health and Nutrition Examination Survey
NNAL	=	4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol
NNK	=	4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone
NS	=	never smokers
PS	=	past smokers
WBC	=	white blood cells.

Acknowledgements

The authors acknowledge the editorial assistance of Eileen Y. Ivasauskas of Accuwrit Inc.

Disclosure statement

QL, RMK, and MS are employees of Altria Client Services LLC and KS was a paid summer intern at Altria Client Services LLC.