Comparison of Biomarkers of Tobacco Exposure between Premium and Discount Brand Cigarette Smokers in the NHANES 2011-2012 Special Sample

Abstract

Background

Increased cigarette costs have inadvertently strengthened the appeal of discounted brands to price-sensitive smokers. While smokers perceive discounted brands as having poorer quality, little is known about their delivery of toxic tobacco smoke constituents compared to premium-branded tobacco products.

Methods

We investigated the differences between discount and premium brand smokers using the National Health and Nutrition Examination Survey (NHANES) 2011-2012 Special Smoker Sample. Our analyses focused on demographic differences and 27 biomarkers of harmful and potentially harmful constituents (HPHCs) listed by the FDA, including volatile organic compounds (VOCs), 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol and its glucuronide (4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol glucuronide) (reported as total NNAL, tNNAL), metals, and polycyclic aromatic hydrocarbons (PAHs). Data were analyzed using linear regression models adjusting for potential confounders.

Results

A total of 976 non-tobacco users and 578 recent cigarette smokers were eligible for analysis, of which 141 (26.0% weighted) smoked discount brand cigarettes and 437 (74.0% weighted) smoked premium. Discount brand smokers were older, predominantly non-Hispanic white, and had higher serum cotinine. Discount brand smokers had significantly higher levels of 13 smoking-related biomarkers, including tNNAL, uranium, styrene, xylene, and biomarkers of exposure to PAHs (naphthalene, fluorene, and phenanthrene), compared to premium brand smokers.

Conclusions

These findings suggest that discount cigarette use is associated with higher exposure to several carcinogenic and toxic HPHCs.

Impact

These results may have important regulatory implications for product standards, as higher exposures could lead to a greater degree of harm.

Introduction

Cigarette smoking still remains the cause of more than 440,000 deaths annually in the U.S. alone (1), which can be reduced by rigorous public health policies. Raising cigarette unit price via taxation has been one of the most effective public health policies that has reduced cigarette consumption, deterred initiation, and encouraged quitting (2–4). However, cigarette companies have counteracted with aggressive pricing strategies to mitigate the impact of these policies (5–7). One such pricing strategy was the introduction of lower priced discount/generic brands in the 1980s to attract price-conscious smokers. Over the years, premium brands have also devised price reduction strategies by offering multipack discounts, rebates, and coupons, in attempts to prevent smokers from switching to more cost-effective brands.

There are several research studies (8–12) and internal tobacco industry documents (13–22) that have investigated usage trends and socio-demographic differences between smokers of generic/discount versus premium brands. Out of 260 cigarette brands on the market in the United States in 2011, 83% of them were classified as discount brands via marketing and advertising, and an increasing number of smokers (30%) used them (8, 23). Industry documents reveal that cigarette companies studied smokers’ behaviors and perceptions for the development of their branding, pricing, and marketing strategies of their discount brands. They found that the primary reason cited for smoking a discount brand was the lower price. Interestingly, research shows that smokers perceived discount brands to be manufactured with poorer quality (strength, satisfaction, draw, pack-to-pack consistency, evenness of burning, tightness and packing of tobacco in cigarette, etc.) and to taste less satisfying (9, 14, 20). Studies indicate that discount brand smokers were more likely to be older, have lower household incomes, and higher nicotine dependence, making them more vulnerable to the health risks of smoking (12, 23).

Surprisingly, to our knowledge, there are no scientific studies that have investigated if American premium and discount brands differ in their manufacturing, toxicant delivery, or potential harm. The National Health and Nutrition Examination Survey (NHANES) (24) provides an opportunity to examine biomarkers of exposure (BOE) to tobacco in a sample of subjects representative of the general U.S. population. NHANES is a comprehensive and publicly available data source that uses a complex, multistage sampling procedure to gather a nationally representative sample of the civilian, noninstitutionalized population (25). NHANES collects data via home surveys, Mobile Examination Center (MEC) visits and questionnaires, specimen collections, and various other physiological and behavioral measurements (26, 27). NHANES began oversampling the smoking population in a Special Sample during the 2011-2012 data collection wave (28).

The primary objective of this study was to examine whether there are differences in BOE concentrations among smokers of premium versus discount cigarette brands using NHANES Special Sample data collected from the 2011-2012 sampling wave. This study investigates the differences between discount and premium brand cigarette user demographics and exposures, in order to assess and compare potential harm on the smoking population.

Materials and Methods

NHANES Special Sample

The National Center for Health Statistics Research Ethics Review Board approved Protocol #2011-17 for NHANES 2011-2012, for which written informed consent was obtained for all participants (24). NHANES 2011-2012 captured data for both smokers and non-smokers. A subsample (1/3 of NHANES participants age 6 or older) was selected for urine collection. The Special Sample consists of all those 20 years or older within the subsample, and all smokers 20 years or older (Figure 1). For these purposes, a smoker was defined by NHANES as having smoked at least 100 cigarettes in their lifetime and now smoking cigarettes every day (28).

Figure 1.

Flowchart of NHANES 2011-2012 Special Smoker Sample.

Figure 1 depicts the NHANES 2011-2012 participants included in the Special Smoker Sample for the analysis of biomarkers of tobacco exposure.

Classification of Non-Tobacco Users and Recent Cigarette Smokers

Participants completed survey questionnaires during an At-Home interview and at a MEC appointment (27). Specimen collection and more recent tobacco-use information was gathered at the MEC (29, 30), therefore giving precedent to MEC-collected data over At-Home collected data for determination of smoking status in the days leading up to specimen collection. Recent cigarette smokers were defined as having used only cigarettes in the last 5 days, while non-tobacco users were defined as either a) smoked more than 100 cigarettes in their life but do not smoke at all now and did not use any tobacco or nicotine products in the last 5 days or b) smoked less than 100 cigarettes in their life and did not use any tobacco or nicotine products in the last 5 days (31). A cutoff of 10 ng/mL of serum cotinine was used as a secondary confirmation for smoking status, with recent cigarette smokers having levels higher than 10 ng/mL and non-tobacco users having levels of 10 ng/mL or less. The 10 ng/mL cutoff is a reasonable accommodation within the range of previously established cutoffs (3-20 ng/mL) (32), with conservative consideration for second-hand smoke exposure. Generally speaking, the smoking status definition from the self-reported survey question items did not deviate much from the cotinine definition. Participants were excluded from analysis if they had not been categorized as either a non-tobacco user or a recent cigarette smoker according to these survey question criteria or if their serum cotinine levels were inconsistent with their self-reported use history. Only 14 of self-reported non-tobacco users had cotinine levels higher than 10 ng/mL and only 16 of self-reported recent cigarette smokers had cotinine levels 10 ng/mL or less. Of these, only 1 reported smoking a discount brand, while the remaining 15 reported smoking a premium brand.

Defining Discount and Premium Brand Smokers

For those identified as recent cigarette smokers, their cigarettes were classified as either a premium brand or a discount brand according to Cornelius et al.’s listed criteria (23), as this is the most comprehensive division of premium and discount brands reported in the literature. Using these criteria, a brand is categorized as a premium brand using the advertising image, leading the consumer to believe it is of higher value than alternative brands of cigarettes. Although discount brands are generally marketed at a lower price, it is the marketing tactic, rather than the price-point, that defines the classification (23). To eliminate the possibility of any misclassifications, and to be consistent with definitions published in the literature, any recent cigarette smokers that reported smoking brands not otherwise categorized by Cornelius et al. (23) were not eligible for analysis. A complete list of reported brand names and their respective categorizations is provided in Supplementary Table S1.

Analysis Population

The NHANES 2011-2012 Special Sample had 2,349 observations. Of these, 77 had non-positive two-year smoking weights and were therefore ignored for all statistical procedures (n=2,272) (33). To be eligible for analysis, participants must have completed the recent tobacco-use survey administered in the MEC, could not have used any other type of tobacco product aside from cigarettes within the last 5 days prior to biosample collection, have been classified as either a non-tobacco user or a recent cigarette smoker, and must have reported non-missing brand information (if categorized as a recent cigarette smoker). Recent cigarette smokers who reported smoking herbal cigarettes or smoking brands that were not defined by Cornelius et al. (23) were not eligible for analysis. After excluding observations with missing data for independent covariates (complete case analysis), a total of 73.0% (weighted, 1554/2272) were eligible for analysis of non-tobacco users versus recent cigarette smokers; of these, 82.2% (weighted, 976/1554) were non-tobacco users and 17.8% (weighted, 578/1554) were recent cigarette smokers. Analysis of discount versus premium brand smokers focused on the subpopulation of 578 recent cigarette smokers, of which 26.0% (weighted, 141/578) smoked discount brands and 74.0% (weighted, 437/578) smoked premium brands.

Analysis Biomarkers

Our focus was on biomarkers of chemicals listed on the FDA’s harmful and potentially harmful constituents (HPHC) list. These included HPHCs or their metabolites (e.g., 1-hydroxypyrene), and any likely metabolites known to be related to HPHCs (e.g., 2-hydroxyfluorene) (34, 35). Using this list, it was then determined which of these HPHCs were available for analysis in the NHANES 2011-2012 Special Sample (36). Serum cotinine and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol and its glucuronide (4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol glucuronide) (reported as total NNAL, tNNAL) were not limited to the Special Sample, but were measured in a larger eligible population of the total NHANES sample (37). If multiple metabolites were available for a specific HPHC, only the major metabolite of the compound was used, as long as there was a well-established precedent in the literature (e.g., acrolein via N-Acetyl-S-(2-carboxyethyl)-L-cysteine and N-Acetyl-S-(3-hydroxypropyl)-L-cysteine (38–45)). If there were no clear delineations, both available metabolites were analyzed (e.g., styrene via mandelic acid and phenylglyoxylic acid). All urinary biomarker concentrations were corrected for dilution by creatinine and all concentrations were reported per gram of creatinine.

If the biomarker concentration was below the specified limit of detection (LOD) for the corresponding metabolite, NHANES sets the default concentration value equal to the limit of detection divided by the square root of two (Supplementary Table S2) (46). Vinyl chloride (12.5% weighted, 65/577), uranium (21.2%, 113/578), and 4-phenanthrene (8.4%, 36/572) had the most smoker samples measuring below the specified limits of detection (Supplementary Table S3). The percent measuring below LOD for vinyl chloride, uranium, and 4-phenanthrene were similar for both discount and premium brand smokers (Supplementary Table S4). The rest of the biomarkers considered had less than 4% of smoker samples measuring below the LOD.

Statistical Analysis

The NHANES 2011-2012 Special Sample of 2,349 observations acted as the least common denominator dataset for all analyses, merging information from the smoking questionnaires and demographics. The Two-Year Smoking Weights (WTFSM) from the Special Sample were used for all analyses, along with the stratification variable (SDMVSTRA) and clustering variable (SDMVPSU) to account for the complex-sampling design of NHANES and the representative weighting of the sampling procedures, which utilized primary sampling units and cluster sampling (47). All reported frequencies are unweighted observations for ease of interpretability, and all reported percentages are weighted and design-corrected.

The following covariates were used to adjust for potential confounders in the regression models: gender, race, age, education, height, weight, serum cotinine, income poverty ratio, and cigarette rod length (for smokers only). Urinary creatinine is known to differ by gender, race, and age, and is influenced by height and weight (48). Race, education, and income poverty ratio are well-established indicators of socioeconomic status, which is associated with smoking status (49, 50). While self-reported measures of cigarettes per day (CPD) and when the participants smoked their last cigarette are relatively standard questionnaire items, they are flawed as proxies for measuring dose exposure. These types of self-reported responses are commonly subject to imprecision and systematic errors, such as digit and recall bias (51). Instead, serum cotinine was used as a more accurate physiological measure of recent nicotine exposure, as opposed to reliance on survey self-report. Additionally, cigarette rod length has been reported to have an effect on biomarker concentrations as well, specifically tNNAL (52).

The relationships between the categorical and continuous descriptive characteristics in Tables 1 and 2 were analyzed using Rao-Scott Chi-Square tests (PROC SURVEYFREQ) and two-sample t-tests (PROC SURVEYREG, noint option, vadjust=none option (47)), respectively. Multivariable linear regression models (PROC SURVEYREG) were used to obtain the covariate-adjusted geometric means for each creatinine-corrected biomarker concentration in Tables 3 and 4, along with the ratio of geometric means (LSMEANS) to assess significant differences in BOE concentrations between study groups. All creatinine-corrected urinary biomarker concentrations were natural-log transformed to better satisfy the normality assumption. SAS SURVEY Procedures of SAS Version 9.4 (SAS Institute Inc., Cary, NC, USA) were used to perform all statistical analyses at a two-sided significance level of 0.05. All procedures implemented the NOMCAR option and the Taylor Series Linearization method for variance estimation, as recommended by NHANES (47, 53). DOMAIN statements were used to perform all subpopulation analyses, which were dependent upon user-defined domains (recent cigarette smokers vs. non-tobacco users, discount vs. premium brand smokers) based on previously outlined analysis eligibility and classification criteria.

Table 1.

Non-Tobacco Users vs. Cigarette Smokers Demographics

	Non-Tobacco User (n=976)	Cigarette Smoker (n=578)	Difference	p-value
Gender			–	0.15
Male	472 (44.9 ± 2.2 [40.3, 49.4])	337 (52.2 ± 3.6 [44.7, 59.7])
Female	504 (55.1 ± 2.2 [50.6, 59.7])	241 (47.8 ± 3.6 [40.3, 55.3])
Race/Ethnicity			–	0.47
Mexican American/Hispanic	197 (14.4 ± 2.8 [8.6, 20.2])	77 (10.5 ± 2.5 [5.2, 15.9])
Non-Hispanic White	362 (68.9 ± 3.7 [61.1, 76.7])	280 (71.5 ± 5.1 [60.8, 82.2])
Non-Hispanic Black	235 (9.6 ± 2.1 [5.2, 14.0])	169 (11.2 ± 2.6 [5.6, 16.7])
Other Race - Including Multi-Racial	182 (7.1 ± 1.0 [5.0, 9.1])	52 (6.8 ± 2.1 [2.4, 11.2])
Education			–	<0.001
Less Than HS	179 (11.2 ± 1.6 [7.8, 14.6])	156 (22.4 ± 1.6 [18.9, 25.8])
HS or Higher	797 (88.8 ± 1.6 [85.4, 92.2])	422 (77.6 ± 1.6 [74.2, 81.1])
Standing Height (cm)	167.8 ± 0.6 (166.6, 168.9)	170.3 ± 0.6 (169.1, 171.5)	2.5 ± 0.8 (0.8, 4.3)	0.01
Weight (kg)	81.7 ± 1.0 (79.7, 83.7)	80.2 ± 1.1 (77.8, 82.5)	−1.5 ± 1.2 (−4.1, 1.0)	0.22
Age (in Years) at Screening	48.7 ± 1.2 (46.2, 51.2)	45.7 ± 0.6 (44.3, 47.0)	−3.1 ± 1.2 (−5.6, −0.5)	0.02
Ratio of Family Income to Poverty	3.1 ± 0.1 (2.9, 3.4)	2.3 ± 0.1 (2.0, 2.6)	−0.8 ± 0.1 (−1.1, −0.5)	<0.001
Serum Cotinine (ng/mL)	0.1 ± 0.0 (0.1, 0.2)	228.4 ± 6.5 (214.7, 242.1)	228.3 ± 6.5 (214.7, 241.9)	<0.001

Categorical Variables: Unweighted Frequencies (Weighted Column % ± Standard Error, [95% CI]), Rao-Scott Chi-Square Adj. F p-value

Continuous Variables: Weighted Mean ± Standard Error (95% CI), Difference Cigarette Smoker – Non-Tobacco User in Least Squared Means (95% CI); Wald p-value

CI = Confidence Interval

Table 2.

Discount vs. Premium Brand Cigarette Smokers Demographics

	Discount (n=141)	Premium (n=437)	Difference	p-value
Gender			–	0.99
Male	81 (52.2 ± 3.6 [44.6, 59.9])	256 (52.2 ± 4.8 [42.0, 62.3])
Female	60 (47.8 ± 3.6 [40.1, 55.4])	181 (47.8 ± 4.8 [37.7, 58.0])
Race/Ethnicity			–	<0.001
Mexican American/Hispanic	5 (2.4 ± 1.1 [0.1, 4.7])	72 (13.4 ± 3.2 [6.6, 20.2])
Non-Hispanic White	102 (81.8 ± 4.4 [72.6, 91.0])	178 (67.9 ± 5.6 [56.0, 79.8])
Non-Hispanic Black	25 (5.9 ± 2.5 [0.7, 11.2])	144 (13.0 ± 2.6 [7.5, 18.5])
Other Race - Including Multi-Racial	9 (9.9 ± 4.2 [1.1, 18.6])	43 (5.7 ± 2.0 [1.6, 9.9])
Education			–	0.53
Less Than HS	43 (25.6 ± 5.1 [14.8, 36.3])	113 (21.2 ± 2.6 [15.8, 26.7])
HS or Higher	98 (74.4 ± 5.1 [63.7, 85.2])	324 (78.8 ± 2.6 [73.3, 84.2])
Cigarette Rod Length			–	0.003
Regular/King (68-72 mm, 79-88 mm)	63 (44.3 ± 6.0 [31.7, 57.0])	263 (65.0 ± 4.4 [55.8, 74.2])
Long/Ultra Long (94-101 mm, 110-121 mm)	78 (55.7 ± 6.0 [43.0, 68.3])	174 (35.0 ± 4.4 [25.8, 44.2])
Cigarette Flavor			–	0.10
Menthol	29 (20.3 ± 3.7 [12.4, 28.1])	181 (30.5 ± 3.0 [24.1, 36.9])
Non-Menthol	112 (79.7 ± 3.7 [71.9, 87.6])	256 (69.5 ± 3.0 [63.1, 75.9])
Cigarette Type			–	0.55
Filtered	138 (98.8 ± 0.9 [96.8, 100.0])	429 (97.8 ± 1.3 [95.1, 100.0])
Non-Filtered	3 (1.2 ± 0.9 [0.0, 3.2])	8 (2.2 ± 1.3 [0.0, 4.9])
When Last Smoked			–	0.04
Today	129 (93.1 ± 2.3 [88.2, 97.9])	371 (81.8 ± 3.5 [74.4, 89.2])
Within the last 5 days (not including today)	12 (6.9 ± 2.3 [2.1, 11.8])	66 (18.2 ± 3.5 [10.8, 25.6])
Number of Days Smoked in the Last 5 Days			–
All 5 Days	129 (92.3 ± 2.6 [86.9, 97.8])	383 (83.2 ± 2.8 [77.3, 89.0])		0.09
1-4 Days	12 (7.7 ± 2.6 [2.2, 13.1])	54 (16.8 ± 2.8 [11.0, 22.7])
Standing Height (cm)	170.4 ± 0.8 (168.8, 172.0)	170.3 ± 0.7 (168.8, 171.8)	0.1 ± 1.0 (-2.0, 2.2)	0.91
Weight (kg)	79.5 ± 2.3 (74.7, 84.3)	80.4 ± 1.5 (77.2, 83.6)	−0.9 ± 3.1 (−7.4, 5.6)	0.77
Age (in Years) at Screening	53.3 ± 1.6 (49.9, 56.7)	43.0 ± 0.5 (41.9, 44.0)	10.4 ± 1.5 (7.3, 13.4)	<0.001
Ratio of Family Income to Poverty	2.3 ± 0.3 (1.7, 2.9)	2.3 ± 0.1 (2.0, 2.6)	0.0 ± 0.3 (-0.6, 0.6)	0.97
Serum Cotinine (ng/mL)	264.0 ± 8.1 (246.9, 281.1)	215.9 ± 9.2 (196.5, 235.3)	48.1 ± 11.7 (23.4, 72.8)	<0.001
CPD in Last 5 Days	17.5 ± 0.9 (15.6, 19.3)	11.7 ± 0.7 (10.2, 13.2)	5.8 ± 1.2 (3.2, 8.3)	<0.001

Categorical Variables: Unweighted Frequencies (Weighted Column % ± Standard Error, [95% CI]), Rao-Scott Chi-Square Adj. F p-value

Continuous Variables: Weighted Mean ± Standard Error (95% CI), Difference Cigarette Smoker – Non-Tobacco User in Least Squared Means (95% CI); Wald p-value

CI = Confidence Interval

Table 3.

Non-Tobacco Users vs. Cigarette Smokers Regression Results

Outcome	Units	Geometric Mean (95% CI)		Ratio of GMs (95% CI)	p-value
		Non-Tobacco Users	Cigarette Smokers
Volatile Organic Compounds
Xylene (2-MHA)	μg/g creatinine	38.3 (32.9, 44.5)	75.8 (64.5, 89.1)	0.5 (0.4, 0.6)	<0.001
Xylene(3-MHA & 4-MHA)	μg/g creatinine	250.9 (221.6, 284.1)	517.0 (437.1, 611.4)	0.5 (0.4, 0.6)	<0.001
Acrylamide	μg/g creatinine	47.3 (42.6, 52.5)	86.3 (75.5, 98.6)	0.5 (0.4, 0.7)	<0.001
Toluene	μg/g creatinine	7.5 (6.5, 8.7)	6.8 (5.5, 8.3)	1.1 (0.8, 1.5)	0.52
Acrylonitrile	μg/g creatinine	2.4 (2.1, 2.8)	87.7 (74.2, 103.8)	0.0 (0.0, 0.0)	<0.001
1,3-Butadiene	μg/g creatinine	281.6 (258.5, 306.7)	340.7 (311.7, 372.5)	0.8 (0.7, 1.0)	0.01
Vinyl Chloride	μg/g creatinine	1.2 (1.1, 1.4)	2.3 (1.9, 2.8)	0.5 (0.4, 0.7)	<0.001
Propylene Oxide	μg/g creatinine	35.1 (30.5, 40.4)	46.5 (39.5, 54.6)	0.8 (0.6, 0.9)	0.02
Acrolein	μg/g creatinine	291.4 (259.1, 327.8)	611.4 (543.7, 687.5)	0.5 (0.4, 0.6)	<0.001
Crotonaldehyde	μg/g creatinine	526.0 (471.1, 587.3)	1281.7 (1086.6, 1512.0)	0.4 (0.3, 0.5)	<0.001
Styrene (Mandelic Acid)	μg/g creatinine	147.4 (136.1, 159.7)	207.7 (183.3, 235.3)	0.7 (0.6, 0.8)	<0.001
Styrene (Phenylglyoxylic Acid)	μg/g creatinine	192.0 (174.1, 211.7)	285.6 (252.6, 323.0)	0.7 (0.6, 0.8)	<0.001
Nitrosamines
Total (t)NNAL	ng/g creatinine	1.6 (1.4, 1.9)	161.1 (129.2, 200.9)	0.0 (0.0, 0.0)	<0.001
Cyanide	μg/g creatinine	948.9 (863.2, 1043.0)	2900.8 (2473.6, 3401.7)	0.3 (0.3, 0.4)	<0.001
Metals
Cobalt	ng/g creatinine	355.2 (319.4, 395.1)	299.9 (259.3, 346.8)	1.2 (1.0, 1.4)	0.08
Lead	ng/g creatinine	429.5 (380.8, 484.3)	544.9 (439.3, 675.7)	0.8 (0.6, 1.0)	0.03
Uranium	ng/g creatinine	6.1 (5.2, 7.1)	7.7 (6.7, 8.8)	0.8 (0.6, 1.0)	0.03
Polycyclic Aromatic Hydrocarbons
1-hydroxynaphthalene	μg/g creatinine	1.8 (1.5, 2.1)	8.6 (6.6, 11.3)	0.2 (0.1, 0.3)	<0.001
2-hydroxynaphthalene	μg/g creatinine	5.2 (4.7, 5.8)	11.6 (10.2, 13.4)	0.4 (0.4, 0.5)	<0.001
3-hydroxyfluorene	ng/g creatinine	93.3 (84.5, 103.1)	428.6 (372.3, 493.4)	0.2 (0.2, 0.3)	<0.001
2-hydroxyfluorene	ng/g creatinine	261.9 (238.0, 288.2)	916.4 (829.5, 1012.3)	0.3 (0.2, 0.3)	<0.001
3-hydroxyphenanthrene	ng/g creatinine	73.1 (63.5, 84.2)	116.5 (94.8, 143.2)	0.6 (0.5, 0.8)	0.005
1-hydroxyphenanthrene	ng/g creatinine	139.0 (126.8, 152.4)	171.8 (153.5, 192.4)	0.8 (0.7, 1.0)	0.01
2-hydroxyphenanthrene	ng/g creatinine	73.8 (62.9, 86.5)	107.6 (86.2, 134.2)	0.7 (0.5, 1.0)	0.04
1-hydroxypyrene	ng/g creatinine	124.5 (110.7, 139.9)	198.9 (170.7, 231.8)	0.6 (0.5, 0.8)	<0.001
9-hydroxyfluorene	ng/g creatinine	283.0 (238.3, 336.1)	569.5 (446.7, 726.2)	0.5 (0.3, 0.7)	0.001
4-phenanthrene	ng/g creatinine	25.4 (21.5, 29.9)	34.3 (27.0, 43.6)	0.7 (0.5, 1.1)	0.10

Geometric Means (GMs) are adjusted for Gender, Age, Income Poverty Ratio, Race, Education, Serum Cotinine (ng/mL), Height (cm), and Weight (kg)

CI = Confidence Interval

Table 4.

Discount vs. Premium Brand Cigarette Smokers Regression Results

Outcome	Units	Geometric Mean (95% CI)		Ratio of GMs (95% CI)	p-value
		Discount	Premium
Volatile Organic Compounds
Xylene (2-MHA)	μg/g creatinine	135.9 (110.2, 167.5)	112.7 (98.8, 128.5)	1.2 (0.9, 1.6)	0.13
Xylene(3-MHA & 4-MHA)	μg/g creatinine	1018.7 (850.4, 1220.2)	805.9 (697.7, 930.9)	1.3 (1.0, 1.5)	0.02
Acrylamide	μg/g creatinine	123.5 (109.3, 139.5)	126.7 (118.8, 135.1)	1.0 (0.9, 1.1)	0.69
Acrylonitrile	μg/g creatinine	168.2 (144.5, 195.7)	144.8 (123.8, 169.2)	1.2 (0.9, 1.4)	0.16
1,3-Butadiene	μg/g creatinine	417.5 (375.5, 464.2)	381.4 (360.4, 403.6)	1.1 (1.0, 1.2)	0.07
Vinyl Chloride	μg/g creatinine	2.9 (2.2, 3.8)	3.2 (2.8, 3.7)	0.9 (0.7, 1.1)	0.31
Propylene Oxide	μg/g creatinine	74.5 (58.2, 95.4)	68.3 (61.2, 76.1)	1.1 (0.8, 1.5)	0.54
Acrolein	μg/g creatinine	1139.9 (996.3, 1304.1)	966.0 (846.5, 1102.5)	1.2 (1.0, 1.4)	0.08
Crotonaldehyde	μg/g creatinine	2458.1 (2116.4, 2855.1)	2115.1 (1833.9, 2439.3)	1.2 (1.0, 1.4)	0.10
Styrene (Mandelic Acid)	μg/g creatinine	312.8 (279.2, 350.4)	256.0 (231.6, 282.9)	1.2 (1.1, 1.4)	0.003
Styrene (Phenylglyoxylic Acid)	μg/g creatinine	396.3 (346.0, 453.9)	331.2 (301.9, 363.4)	1.2 (1.0, 1.4)	0.04
Nitrosamines
Total (t)NNAL	ng/g creatinine	308.5 (261.2, 364.2)	254.7 (217.5, 298.2)	1.2 (1.0, 1.4)	0.04
Cyanide	μg/g creatinine	4403.9 (3304.3, 5869.4)	4570.3 (4045.9, 5162.7)	1.0 (0.7, 1.4)	0.83
Metals
Lead	ng/g creatinine	535.3 (385.6, 743.2)	514.6 (410.2, 645.6)	1.0 (0.8, 1.3)	0.72
Uranium	ng/g creatinine	9.5 (7.7, 11.9)	7.2 (6.2, 8.2)	1.3 (1.1, 1.7)	0.01
Polycyclic Aromatic Hydrocarbons
1-hydroxynaphthalene	μg/g creatinine	14.2 (11.8, 17.0)	10.6 (9.8, 11.4)	1.3 (1.1, 1.7)	0.01
2-hydroxynaphthalene	μg/g creatinine	17.9 (15.8, 20.1)	14.2 (13.0, 15.5)	1.3 (1.1, 1.5)	0.004
3-hydroxyfluorene	ng/g creatinine	855.9 (764.8, 958.0)	693.1 (632.6, 759.5)	1.2 (1.1, 1.4)	0.01
2-hydroxyfluorene	ng/g creatinine	1640.5 (1442.0, 1866.3)	1315.3 (1199.0, 1442.8)	1.2 (1.0, 1.5)	0.02
3-hydroxyphenanthrene	ng/g creatinine	198.3 (165.2, 238.0)	161.6 (145.0, 180.1)	1.2 (1.0, 1.4)	0.01
1-hydroxyphenanthrene	ng/g creatinine	251.8 (222.7, 284.7)	209.8 (192.7, 228.4)	1.2 (1.0, 1.4)	0.01
2-hydroxyphenanthrene	ng/g creatinine	150.4 (131.7, 171.8)	125.1 (113.3, 138.1)	1.2 (1.0, 1.4)	0.03
1-hydroxypyrene	ng/g creatinine	305.3 (264.8, 352.0)	272.4 (242.3, 306.2)	1.1 (1.0, 1.3)	0.06
9-hydroxyfluorene	ng/g creatinine	952.4 (824.7, 1099.8)	681.8 (609.3, 763.0)	1.4 (1.2, 1.6)	<0.001

Geometric Means (GMs) are adjusted for Gender, Age, Income Poverty Ratio, Race, Education, Serum Cotinine (ng/mL), Height (cm), Weight (kg), and Rod Length

CI = Confidence Interval

Results

Demographics

Prior to evaluating differences in smokers according to the type of brand that they smoke, it is important to discern a baseline standard of differences that exist between recent cigarette smokers and non-tobacco users overall. Table 1 shows that 88.8% (weighted, 797/976) of non-tobacco users had at least a high school degree versus only 77.6% (weighted, 422/578) of cigarette smokers. Non-tobacco users tended to be about 3 years older, had a higher family income to poverty ratio, and had much lower cotinine levels, than cigarette smokers. In addition to these differences, which have been consistently reported in prior publications (54–56), there was also an observed difference in standing height, with smokers being about 2 cm taller than non-smokers. While the difference is statistically significant, the size of the effect is negligible and can be explained by the slightly larger proportion of males characterized as recent cigarette smokers (52.2% vs. 44.9%). Exploring the demographic characteristics of discount and premium brand cigarette smokers revealed some interesting differences between the users (Table 2). Discount smokers were, on average, about ten years older and mostly non-Hispanic white (81.8% weighted, 102/141). Due to the observed difference in age, we further analyzed whether the two groups of users differed with respect to the age at which they began smoking regularly, as it would be more informative to look at how long, on average, each of the cohorts had been smoking. The mean age that discount brand smokers started smoking regularly (17.9 years) did not differ (p=0.65) from premium brand smokers (17.5 years). Therefore, age at the time of screening can be viewed as an effective proxy for how long each group of smokers had been smoking cigarettes regularly (discount: 35.4 years, premium: 25.5 years). In addition, discount brand smokers had significantly higher serum cotinine (264.0 ng/mL vs. 215.9 ng/mL), were significantly more likely to have smoked on the same day as their MEC visit (93.1% vs. 81.8%), and reported smoking significantly more cigarettes per day in the last 5 days (18 CPD vs. 12 CPD) compared to premium brand cigarette smokers. Because of this, we further explored whether the two groups of users differed with respect to their dependence on the products. Using their reported time to first cigarette (TTFC) (57), it was found that the two groups did not differ in their dependence on cigarettes (p=0.46). However, TTFC was only collected for participants that self-reported using cigarettes every day; 97.8% (139/141) of discount brand users reported daily use compared to 84.2% (413/437) of premium brand users (p=0.03). While self-reported daily use may have significantly differed between the users, the actual number of days that each group used cigarettes leading up to their biosample collection did not significantly differ, with 92.3% (129/141) of discount users using all 5 days prior to their MEC visit compared to 83.2% (383/437) of premium users (p=0.09). Because of these slight inconsistencies in recent smoking, we used cotinine as the controlling covariate for nicotine exposure in the regression models. The two groups of product users did not significantly differ with respect to the filter or menthol status of their cigarettes. This was expected, as most cigarettes on the market are filtered, while most, but not all, brands are available in both menthol and non-menthol varieties.

Biomarker Concentrations

We first compared 27 BOEs in smokers versus non-smokers (Table 3) as a conservative way of evaluating the discerning abilities of the reported biomarkers to characterize general toxicant exposure, as compared with tobacco smoke exposure. If exposures significantly differed between the two groups, then there was a biological rationale for then evaluating whether the exposure also differed with respect to the types of products the smokers used. Only three biomarkers did not significantly differ between smokers and non-smokers: toluene, cobalt, and 4-phenanthrene. The remaining 24 biomarkers were all significantly higher in smokers than non-smokers, with differences most marked for acrylonitrile and tNNAL.

The 24 biomarkers that were significantly higher in smokers provided the basis for regression analyses examining differences between premium and discount brand smokers (Table 4). Concentrations of 13 of the 24 biomarkers were significantly higher in discount brand smokers compared to premium smokers: xylene (3-MHA + 4-MHA), styrene (mandelic acid and phenylglyoxylic acid), tNNAL, uranium, 1- and 2-hydroxynaphthalene, 9-, 3-, and 2-hydroxyfluorene, and 3-, 2-, and 1-hydroxyphenanthrene.

While the remaining eleven biomarkers did not significantly differ between the two types of users, most (8/11) appeared to have higher levels in discount brand smokers, such as xylene (2-MHA), acrylonitrile, 1,3-butadiene, propylene oxide, acrolein, crotonaldehyde, lead, and 1-hydroxypyrene. Only three biomarkers had slightly higher (but not significantly different) concentrations in premium brand smokers: acrylamide, vinyl chloride, and cyanide.

Discussion

Through examination of a nationally representative sample, we have explored inherent characteristic differences, along with differences in biomarkers of tobacco exposure, between premium and discount brand smokers. BOE of tobacco constituents/toxicants are metabolites that are considered measures of internal toxicant dose. Thus, through better understanding of differences and variation in BOE levels among smokers of different types of cigarette products, regulatory science can be better informed about cigarette characteristics associated with higher toxicant exposure.

Reaffirming the national representativeness of the sample, the proportion of cigarette tobacco users in NHANES agrees with the national averages of smokers and non-smokers reported by the CDC and others (54, 56), as does the relative proportion of discount brand users (26.0% weighted, 141/578) (8, 23). We found that smokers had significantly higher concentrations for 24 of the 27 biomarkers examined (Table 3), which generally agrees with the CDC’s Fourth Exposure Report (28). Of the 24 biomarkers that were observed to be elevated in smokers, 13 were significantly higher in smokers that reported smoking discount brand cigarettes. Seven of these biomarkers are BOEs for carcinogens [tNNAL (NNK), mandelic and phenylglyoxylic acid (styrene), 1-hydroxynaphthalene and 2-hydroxynaphthalene (naphthalene), 1-hydroxypyrene (different pyrenes), and uranium] (34). Of the remaining six, five are known to be strongly correlated with other carcinogens, specifically naphthalene and pyrenes [fluorene (2-hydroxyfluorene, 3-hydroxyfluorene, 9-hydroxyfluorene) and phenathrene (3-hydroxyphenanthrene and 1-hydroxyphenanthrene)] (35). The remaining biomarker, xylene, is a respiratory and nervous irritant (58). With these established biological effects, the differences in BOE between discount and premium brand smokers raise important questions in regards to the health outcomes of smokers using these products, along with their causes. Although the driving forces behind the observed differences remain unclear, we postulate the possibilities of product differences, user differences, and product–user differences.

It may be possible that there are manufacturing, processing, and design differences between products that could lead to elevated levels of toxicant exposures we have observed in smokers using discount brand cigarettes. Our results show that, overall, almost all BOE analyzed (except acrylamide, vinyl chloride, and cyanide), trend higher in those smoking discount brands. Therefore, a difference in total smoke exposure between users of the two products, even after adjusting for cotinine, remains a possibility. One possible explanation could be that discount and premium brand tobacco cigarettes are manufactured differently. Some of these toxicants are present in the tobacco (e.g., uranium), and some are generated during the burning process (e.g., naphthalene). In addition, some additives have been associated with differences in delivery of some toxicants (e.g., sugars and aldehydes) (59). Also, any changes to the cigarettes’ designs could lead to differences in BOE levels. For example, different filter compositions could influence toxicant deliveries to smokers. One study has shown that Canadian discount and premium brand cigarettes share similar compositions, deliveries, and emissions (60). However, to our knowledge, no such analysis has been performed on American discount versus premium brand cigarettes. While the exploration of differences amongst specific brands may at first seem appealing to many researchers, defining what encompasses a ‘brand’ is quite a daunting task. Cigarettes are classified according to brand, color, style, flavor, filter, size, and pack type; thus, matching cigarettes by all of these criteria can be limited by the quality and quantity of the data in any particular study, and the number of subjects smoking any given cigarette can very often be quite small, leading to sample size constraints. For these reasons, evaluating groups of brands, such as premium versus discount, is a more practical approach. Additionally, we were not able to investigate differences in filter ventilation with the data available for this analysis, although we believe this is a potentially important angle to pursue in future studies of such comparisons.

We are not advocating that premium brand cigarettes are in any way ‘safer’ or more ‘safely made’ than discount brand cigarettes. Every cigarette exposes smokers to thousands of toxicants that lead to disease, regardless of the brand being smoked. Dose-response relations, specific toxins, pathogenic mechanisms, and the inter-relationships between the many components of tobacco smoke are not understood with enough confidence to make scientifically reliable quantitative judgments about the risks or actual harm differences amongst brands. This is further emphasized by the fact that eleven BOEs, which could also greatly affect health outcomes, were not observed to be significantly different between the two types of cigarette smokers. While BOEs strongly reflect quantitative magnitudes of tobacco toxicant exposure, only a few have been found to be directly predictive of cancer incidence (61). Yet, the findings of this analysis suggest that possible nuances and differences in manufacturing or ‘branding’ could direct regulatory science to an area in need of much more research and investigation.

Alternatively, since significant differences were not observed among all of the biomarkers that were analyzed, unaccounted differences between the users of the two types of products, such as differences in lifestyle, diet, environmental exposure, secondhand smoke exposure, genetics, metabolism, or other pre-existing health conditions, could also be factors contributing to the differences in BOE levels. The discount brand smokers were predominantly non-Hispanic white and older in age. However, the observed differences in demographics were accounted for in the regression analyses and we still observed meaningful elevations with respect to several PAHs, but hardly any VOCs. While some of the compounds analyzed are highly concentrated in tobacco smoke, many of them are indicative of exposure to multiple sources and are not limited to tobacco exposure. For example, PAHs are known to result from dietary and environmental combustion exposure (62, 63). VOC levels can also be influenced by environmental exposures, such as household cleaners, disinfectant products, and vehicle exhaust (64).

It is also worth noting that smoker perceptions of the brands of cigarettes that they smoke may not follow the logic with which these types of products are marketed (23). Interestingly, in our analysis, we found that the family income to poverty ratio did not significantly differ between the two types of users as expected from previous literature (12, 23), despite the perceived cost and pricing differences in the marketing of the cigarette products. While this was unexpected, the sensitivity and accuracy of self-reported income is worth taking into consideration. Premium brands undoubtedly make up roughly 75% of the market share (8, 23), which explains why so many more smokers were observed to be smoking premium brands, despite almost equivalent income to poverty ratios.

In addition to the aforementioned product and user differences, it is possible that the ways in which these products are being chosen and used by the two populations are also different. We found that discount brand smokers were more likely to be smoking brands with longer cigarette rod lengths, were more likely to have smoked on the same day as their specimen collection, smoked more cigarettes per day in the five days leading up to their MEC visit, and had higher serum cotinine. Thus, it is possible that the preferred design and overall product sensations of the two groups could lead to variations in smoking behavior, and expose the smoker to toxicants and constituents in many ways. Differences in smoking topography have already been established to be related to cotinine (65); therefore, it is likely to affect the BOEs measured in our study as well. Exploring this preliminary hypothesis, we also investigated differences in the amount of serum cotinine per cigarette for each group of users, on average. The geometric mean cotinine-per-cigarette for discount brand smokers was 15.9 ng/ml (95% CI [14.5, 17.5]) compared to 19.3 (95% CI [18.1, 20.6]) for premium brand smokers (p=0.005). This observationally suggests that, despite the evidence supporting discount brand users being heavier smokers and having higher levels of tNNAL and other BOEs, the premium brand smokers are actually exposing themselves to higher levels of nicotine per cigarette. This is all the more surprising when coupled with the prior mention that discount brand users were more likely to smoke brands of longer rod lengths (Table 2); however, since rod lengths are a feature of brand, the association becomes difficult to untangle. Future studies relating smoking topography to BOEs are needed in order to further investigate potential differences in product-use.

Regardless of whether the variation in exposure is due to the products themselves, the differential user demographics, or the ways in which the products are being used, the higher observed BOE levels in discount brand cigarette smokers (even after controlling for key smoker characteristics and cotinine) are suggestive of an increase in potential harm, relative to premium brand smokers. The varying levels of BOEs observed could contribute to explanations of why smokers have such wide ranges of health outcomes. Our results provide a platform for the FDA to begin establishing product standards that could help lower BOEs and reduce potential harms that put all smokers at risk. Tobacco manufacturers are required to report the levels of some HPHCs in their tobacco products (66); thus, future comparisons using these reported data could help determine if our findings may be attributed to characteristics of the products themselves or to how they are being used by their consumers.

Abbreviations List

BOE

Biomarker of Exposure

CDC

Centers for Disease Control

CPD

Cigarettes Per Day

FDA

Food and Drug Administration

HPHC

Harmful and Potentially Harmful Constituent

LOD

Limit of Detection

MEC

Mobile Examination Center

NHANES

National Health and Nutrition Examination Survey

PAH

Polycyclic Aromatic Hydrocarbon

TTFC

Time to First Cigarette

tNNAL

Total NNAL

VOC

Volatile Organic Compound