Urinary Leukotriene E₄ and 2,3-Dinor Thromboxane B₂ Are Biomarkers of Potential Harm in Short-Term Tobacco Switching Studies

Cigarette smoking is an independent risk factor for lung cancer, chronic obstructive pulmonary disease (COPD), and cardiovascular disease (1–4). Cigarette smoke is a dynamic aerosol containing several thousand chemicals, including various toxicants, which are generated during the combustion process. Many of the toxicants cause long-term adverse health outcomes including cancer due to chronic smoking (5). Ninety-three of the toxicants have been designated as harmful and potentially harmful constituents by the FDA (6). Smoking abstinence is the best option to reduce harm from cigarette smoking (7).

Epidemiologic outcomes require the availability of potentially reduced harm products in the marketplace and sustained exclusive use of these products over many years. Hence, interim measures, such as short-term biomarkers, are useful to evaluate the effect of alternate, novel, potentially reduced harm tobacco products on consumers. In the context of tobacco products, such biomarkers have been described as biomarkers of potential harm (BoPH; ref. 8). Several BoPH have been suggested and extensively investigated in smokers (9–11). These markers serve as early indicators of physiologic changes due to product use, which could potentially inform of perturbations in biological processes leading to smoking-related diseases.

Smoking-induced oxidative stress and inflammation are important drivers of underlying disease mechanisms. However, there are few well-established and validated functional BoPH that serve as predictive biomarkers of smoking-related diseases. For example, forced expiratory volume 1 (FEV1) is the most widely used BoPH of respiratory function and is used as a quantitative measure to characterize COPD. Two potential BoPH related to these mechanisms that have consistently distinguished smokers from nonsmokers include F2 isoprostane iP2FαIII and white blood cell counts (12–14).

Combustion-related toxicants drive the adverse health effects associated with cigarette smoking (5, 8, 15). In addition to cigarettes, noncombustible tobacco products including smokeless tobacco products (e.g., chewing tobacco, moist snuff, snus), electronic nicotine delivery systems (ENDS), and tobacco heating products (THP), which do not generate combustion-related toxicants, exist in the current marketplace (16, 17). Although, there are differences in the product constituents and consumer population (18), existing U.S. and Swedish epidemiologic data demonstrate that both products may present less risk than cigarette smoking (19). For example, the risks of lung cancer are much lower for U.S. (11.7-fold) and Swedish (12.8-fold) smokeless tobacco users as well as for switchers from cigarettes to smokeless tobacco than for cigarette smokers (19, 20). Therefore, smokeless tobacco products may serve as an alternative for those smokers who cannot or are unwilling to quit tobacco product use (19, 21). However, there is limited information on the effects of the use of other noncombustible tobacco products (i.e., ENDS and THPs), which produce an aerosol that is chemically far less complex than cigarette smoke.

To better understand the biological and pathophysiologic effects of combustible and noncombustible tobacco products, several cross-sectional biomarker discovery studies, which included cigarette smokers (SMK), moist snuff consumers (MSC), and nontobacco consumers (NTC) have been conducted to identify BoPH for product evaluation (22, 23). Furthermore, SMK exhibit enhanced arachidonic acid (AA) metabolism compared with MSC and NTC, as evidenced by increased AA production (22). In this study, increased production of AA metabolites, isoprostanes and leukotriene E4 (LTE₄), which are markers of oxidative stress and inflammation, were observed. Thus, cigarette smoking evokes a proinflammatory phenotype, highlighted by increased synthesis of AA and its metabolites that include prostaglandins, prostacyclins, thromboxanes, leukotrienes, and hydroxyeicosatetraenoic acids (12, 13, 24).

The biological effects of smoking have been known to persist for a long time after complete cessation, and the evaluation of certain effects of switching to modified risk tobacco products (MRTP) also could require several months. For example, 6 or more months of smoking abstinence is necessary to detect changes in WBC levels (25, 26). A significant change in FEV1 is detectable more than 6–12 months after smoking abstinence (27), suggesting that FEV1 is a long-term biomarker of lung function. Therefore, evaluation of the health effects of potential MRTPs in clinical trials becomes challenging, as confining study volunteers for extended periods to ensure study compliance in residential settings is not practical. Hence, we sought to identify BoPH that rapidly change after a few days of smoking abstinence and/or switching to alternate tobacco products. On the basis of previous studies, which reported that select AA metabolites rapidly change upon smoking abstinence (28), and their established role as markers of oxidative stress and inflammation, we set out to determine whether the AA metabolites would serve as short-term reversible BoPH.

In this study, we assessed the levels of a panel of urinary AA-derived metabolites, including prostaglandins, isoprostanes, thromboxanes, and leukotrienes, as BoPH in smokers who either abstained from smoking or switched to an alternate tobacco product for five days in a residential setting. Urine samples were obtained from three separate short-term product switching studies in which SMK were switched from their usual brand (UB) combustible cigarette to noncombustible products (i.e., oral smokeless tobacco products or ENDS products) or abstinence.

Clinical studies were performed in accordance with the US Code of Federal Regulations (CFR) governing Protection of Human Subjects (21 CFR Part 50), Financial Disclosure by Clinical Investigators (21 CFR Part 54), and Institutional Review Board (IRB; 21 CFR Part 56). In addition to these federal regulations, these studies followed the 1996 guidelines of the International Conference on Harmonization, commonly known as Good Clinical Practice (GCP), which are consistent with the Declaration of Helsinki as adopted in 2008.

A brief description of three sponsored clinical studies [from RJ Reynolds Tobacco Company (RJRT) and RJR Vapor (RJRV) company] in which smokers who completely switched to either abstinence or to test products in a confinement setting for a period of 5 days is provided (Supplementary Fig. S1). The demographics of enrolled subjects in the three studies are summarized in Table 1. The three studies included generally healthy adult male and female smokers who were primarily Caucasian and African American. The mean subject age ranged from 38 to 43 years. The representation of Hispanics was very limited in the study groups.

Study I [Cigarette Per Day Reduction (CPDR) study] was a single-center, randomized, controlled, open-label, parallel group study, designed to evaluate the dose–effect relationships between cigarette per day reduction and biomarkers of exposure (BoE; ref. 29). For the current analysis, 24-hour urine samples from smokers switched from 20 cigarettes per day (CPD) to abstinence (0 CPD cohort) were used to evaluate AA metabolites.

Study II [Modern Smokeless Tobacco Product (STP) study] was a multicenter, open-label, randomized, forced-switching, parallel cohort study designed to evaluate changes in tobacco product use behavior and levels of selected BoE (30). Smokers were randomized into one of six different test product groups: Group 1: dual use of UB cigarettes and Camel Snus; Group 2: exclusive Camel Snus use; Group 3: exclusive Camel Sticks use; Group 4: exclusive Camel Strips use; Group 5: exclusive Camel Orbs use; and Group 6: tobacco product abstinence. In this trial, two Snus variants were available for participant selection: frost (mint) or Mellow (non-mint). The Sticks, Strips, and Orbs tested were generally similar in composition, consisting of finely milled tobacco and other food-grade ingredients, and contained a total alkaloid content of approximately 3.2, 1.0, and 1.1 mg, respectively. These products are used orally and allowed to dissolve in the mouth.

Study III (ENDS exposure study) was a single-center, randomized, controlled, switching, open-label, parallel cohort study (31). Smokers were enrolled and randomized to one of two cohorts switching to either Vuse (VS) Original or Menthol variant, for 5 days after baseline ad libitum smoking of their UB cigarettes. Smoking status at enrollment was defined as self-reported smoking of at least 10 CPD for at least 6 months with an expired carbon monoxide (ECO) level of >15 ppm.

The VS products are composed of a battery, heating element, microchip, sensor, and a cartridge containing propylene glycol, glycerin, nicotine, flavorings, and water. Drawing on the mouthpiece causes the heating element to aerosolize the liquid in the cartridge and delivers a puff of aerosol to the user.

Twenty-four–hour urine samples were collected beginning on study days −3 and 4 (29–31). LC/MS-MS was used to measure urinary AA metabolites including PGF_2α, 8-i-PGF_2α, 2,3-d-8-i-PGF_2α, t-PGDM, t-PGEM, 2,3-d-TXB₂, 11-d-TXB₂, LTE₄, and 12(S)-HETE at Analytisch-Biologisches Forschungslabor GmbH (ABF GmbH; ref. 32).

A biomarker-based gradient boosting model was constructed to classify smoker status (i.e., pre- and postswitching). The model performance, defined as the classifier's ability to identify the tobacco use groups correctly, was computed as the AUC from the receiver operator characteristic (ROC) curve. The ROC curve was plotted as a true positive rate (sensitivity) as a function of false positive rate (1−specificity). To build biomarker-based classifiers for evaluation of VS products, the datasets from Study III (i.e., VS Original and VS Menthol; 75 subjects and 2 time points) were used to build, train, and cross-validate the model. During model fitting, the datasets were randomly split into five folds, four of which were used to train the model and one of which was used for cross-validation. This process (five-fold cross-validation) was repeated 10 times and the average of the model performance metrics were computed.

After model cross-validation, the datasets from Study I (0 CPD cohort) and Study II (Group 6: abstinence cohort) were used as inputs to test the model's accuracy in distinguishing between smokers pre- and postabstinence. The performance of the classification model in the smokeless tobacco product–switching studies was evaluated using the datasets from Groups 2–5 (snus, Sticks, Strips, and Orbs, respectively) in Study II. The “caret” package written in R (http://topepo.github.io/caret/index.html) was used for model training, cross-validation, and prediction.

The BoPH levels are presented as total mass (ng/24 hours). Mean and SD of the biomarkers were calculated for baseline and postswitching in each tobacco use group. To compare the differences between baseline and postswitching in each group, a paired t test was conducted to determine statistical significance (P < 0.05). JMP 10 (SAS Institute) was used for data analysis.

Clinical trials that evaluated the impact of reduced CPD on cigarette smoke toxicant exposure have previously reported reductions in tobacco-related BoE (33, 34). Using well-established BoE, Theophilus and colleagues expanded the tobacco exposure–response relationship by evaluating multiple CPDR levels, including abstinence, in a short-term (5-day) switching study (29). In our effort to identify short-term BoPH, we determined the levels of a panel of AA metabolites in the urine sample collected in the CPDR study.

When compared with baseline, urinary LTE₄ levels 5 days after smoking abstinence were reduced approximately 40% (154 ± 128 vs. 93.1 ± 49.0 ng/24 hours, P = 0.004; Fig. 1). Conversely, following 5 days of abstinence in Study 1, urinary levels of PGF2_α (2,262 ± 975.4 vs. 2,780 ± 1186 ng/24 hours, P = 0.004), 2,3-d-8-i-PGF2_α (4,084 ± 2,394 vs. 4,709 ± 1,380 ng/24 hours, P = 0.02), t-PGEM (15,282 ± 10,927 vs. 16,043 ± 7,415.0 ng/24 hours, P = 0.023), and t-PGDM (3,100 ± 1,274 vs. 4,565 ± 1,646 ng/24 hours, P = 0.0001) increased (Fig. 1). Urinary levels of 11-dehydro-TXB₂, 8-i-PGF2_α, and 2,3-d-TXB₂ were unaltered five days after smoking abstinence.

To determine whether AA metabolism is altered following short-term switching, we measured AA-derived metabolites in SMK switching from UB cigarettes to abstinence, to modern smokeless tobacco products (snus, Sticks, Strips, or Orbs), or to dual use (UB cigarettes and snus) for five days.

Urinary LTE₄ levels were reduced approximately 26%–43% in all cohorts who switched from combustible cigarettes to snus (122 ± 61.2 vs. 68.9 ± 27.9 ng/24 hours, P < 0.001), Sticks (129 ± 101 vs. 95.1 ± 72.4 ng/24 hours, P = < 0.001), Strips (105 ± 37.4 vs. 68.7 ± 34.0 ng/24 hours, P < 0.001), Orbs (125 ± 61.4 vs. 77.0 ± 28.7 ng/24 hours, P < 0.001), or abstinence (155 ± 96.3 vs. 94.0 ± 57.5 ng/24 hours, P < 0.001; Fig. 2). However, urinary LTE₄ levels remained unchanged in subjects switched to dual use of cigarettes and snus (134 ± 78.5 vs. 127 ± 106 ng/24 hours, P > 0.05). Urinary levels of PGF-_2α, 8-iPGF-_2α, 2,3-d-8-iPGF-_2α, t-PGEM, and t-PGDM were unchanged by any switching or forced abstinence.

Urinary levels of 2,3-d-TXB₂ were reduced approximately 24%–48% across all cohorts randomized to short-term switching, including dual use (2,290 ± 1,606 vs. 1,500 ± 1,119 ng/24 hours, P = 0.003), Orbs (1,451 ± 899.7 vs. 832.6 ± 439.7 ng/24 hours, P < 0.001), snus (2,120 ± 2,039 vs. 1,286 ± 1,125 ng/24 hours, P = 0.014), Sticks (1,691 ± 1,000 vs. 1,279 ± 840.9 ng/24 hours, P = 0.037), Strips (1,663 ± 913.8 vs. 923.0 ± 607.2 ng/24 hours, P < 0.001), and abstinence (1,834 ± 921.5 vs. 947.9 ± 577.2 ng/24 hours, P < 0.001; Fig. 3). Urinary 11-dh-TXB₂ levels were reduced in smokers who switched to snus (787 ± 276 vs. 638 ± 207 ng/24 hours, P = 0.009). However, urinary 11-dh-TXB₂ levels were unaltered when subjects switched to Orbs, Sticks, Strips, or abstinence.

Studies have demonstrated BoE reductions in smokers switched to ENDS (35, 36); however, to date, none of these studies evaluated the impact of switching on candidate BoPH. Thus, we sought to identify potential urinary BoPH in smokers switched to either VS Original or VS Menthol. Five days after switching to VS Menthol, urinary levels of PGF-_2α (2,696 ± 1,293 vs. 3,185 ± 1,502 ng/24 hours, P = 0.0009), 8-iPGF-_2α (669 ± 276 vs. 744 ± 268 ng/24 hours, P = 0.0205), and t-PGDM (4,243 ± 2,034 vs. 4,966 ± 1,956 ng/24 hours, P = 0.01413) increased. 2,3-d-TXB₂ urinary levels decreased in smokers switched to either VS Menthol (3,604 ± 1,917 vs. 2,011 ± 1,762 ng/24 hours, P < 0.001) or VS Original (2,561 ± 1,618 vs. 1,439 ± 1,001 ng/24 hours, P < 0.001; Fig. 4). None of the remaining urinary AA metabolites were altered upon switching to either VS Original or VS Menthol.

To assess the clinical validity of the biomarkers, a two-biomarker-based classification model was constructed, and its performance quantified using the AUC from ROC (Fig. 5). In this model, only LTE₄ and 2,3-d-TXB₂ were included, because their urinary levels were significantly decreased across all the various short-term product-switching study cohorts for at least one of the metabolites. The highest AUC was for VS Original (0.88), followed by VS Menthol (0.82) and smoking abstinence (0.72). Furthermore, application of our model using the smokeless tobacco–switching cohorts as input data revealed AUCs of 0.76 (Orbs); 0.61 (Sticks); and 0.72 (snus and Strips; Supplementary Fig. S2).

Chronic smoking is associated with elevated oxidative stress and inflammation, which are key drivers of smoking-induced pathophysiology (13). While several potential BoPH for smoking-related diseases have been identified (11), it usually requires several months of smoking abstinence to detect meaningful changes in their levels. Our goal was to identify and qualify BoPH following short-term smoking abstinence or switching to potential MRTPs. In this article, we investigated whether the AA metabolites are responsive to short-term smoking abstention and whether they can be used as potential BoPH for potential evaluation of tobacco products.

The current work evaluated a panel of known biomarkers of oxidative stress, inflammation, and platelet activation in the urine samples collected from three RJRT and RJRV sponsored studies. Key findings from this study are: (i) switching to either abstinence, dual use of combustible and noncombustible products, or exclusive use of noncombustible products resulted in reduced 2,3-d-TXB₂; and (ii) switching smokers to either abstinence or noninhaled oral tobacco products culminated in reduced LTE₄. Together, these data show 2,3-d-TXB₂ and LTE₄, as potential BoPH. Furthermore, we constructed a classification model based on these BoPH to evaluate their combined performance in distinguishing smokers pre- and postswitching.

The two-biomarker-based model demonstrated the highest performance in distinguishing smokers pre- and postswitching to noncombustible inhaled products (i.e., VS Original and VS Menthol). Moreover, the model performed well (AUC > 0.70) in differentiating smokers switched to either abstinence or smokeless oral tobacco products. Together, these data suggest that a classification model comprising both LTE₄ and 2,3-d-TXB₂ has the potential to categorize smokers subjected to short-term product switching.

Smoking abstinence is the accepted “gold standard” strategy for reducing harm from cigarette smoking. Consequently, the biological effects of switching to a MRTP must be considered in the context of both smoking and abstinence. In Study I, which was designed to evaluate the dose–effect relationships between CPDR and BoE, we found that abstinence resulted in a significant decrease in urinary LTE₄ levels. Leukotrienes (LT) are lipid signaling mediators produced by mast cells, eosinophils, neutrophils, basophils, and macrophages. LT synthesis is initiated by phospholipase A–mediated cleavage of AA, which then undergoes rapid conversion to LTA₄ via 5-lipoxygenase (5-LO) and 5-lipoxygenase-activating protein (FLAP; refs. 37, 38). LTA₄ is subsequently converted to LTC₄, LTD₄, and LTE₄, collectively known as cysteinyl leukotrienes (CysLT), and LTB₄ (Supplementary Fig. S6). Urinary LTE₄, the most stable CysLT, is frequently used as a marker of leukotriene synthesis (39, 40).

Urinary LTE₄ levels are closely correlated with number of cigarettes smoked (41, 42) and cotinine levels (43). In addition, adult smokers had up to 5-fold higher levels of urinary LTE₄ compared with nonsmokers (44). Thus, our data demonstrating increased LTE₄ levels in smokers is consistent with studies examining the link between cigarette smoking and increased LTE₄ levels. Following forced abstinence or switching to smokeless tobacco products, urinary LTE₄ levels declined rapidly and to a similar degree to the decrease seen in smokers reducing the number of cigarettes smoked per day. Our findings align with studies which demonstrate that urinary LTE₄ levels return to baseline within two weeks of tobacco product abstention (42). There was a similar trend toward reduction of urinary LTE₄ levels in smokers switched to either VS Original or VS Menthol; however, the reductions were not statistically significant.

Thromboxane A2 (TXA₂), the major product of prostaglandin endoperoxides in platelets, induces irreversible platelet aggregation (45). TXA₂ is rapidly converted to TXB₂ followed by catabolism to 2,3-d-TXB₂ (46). 2,3-d-TXB₂ is the most abundant metabolite recovered in the urine following intravenous infusion of TXB₂ to humans (47). Thromboxane urinary metabolites are frequently used as a marker for platelet activation in smokers (13, 48–50). For example, 2,3-d-TXB₂ is significantly elevated in healthy chronic smokers when compared with nonsmokers (50). Consistent with the literature, we observed that 2,3-d-TXB₂ urinary levels were reduced following smoking abstinence. Similar results were obtained in smokers switched to dual use, and in smokers switched to exclusive use of noncombustible products (i.e., oral tobacco products and VS products). Together, these data suggest diminished TXA₂ synthesis and platelet activation when smokers switch to noncombustible products.

The lung is a major site of LT synthesis (51) and elevated LT levels in the lung are a hallmark of airway hypersensitivity-related diseases such as asthma and allergic rhinitis. A decrease in LTE₄ levels in smokers who stop smoking or switch to smokeless tobacco products such as snus indicates a decrease in airway hypersensitivity. In smokers who switched to VS Original or VS Menthol, LTE₄ levels were lower (17% and 20%, respectively), although not statistically significant. However, a boosted tree–based classification model based on LTE₄ and 2,3-d-TXB₂ data from all three clinical studies revealed that these two BoPH clearly differentiate smokers who switched to noncombustible products in several days from their baseline levels (Fig. 5). Thus, the levels of LTE₄ and 2,3-d-TXB₂ decline in smokers who switch to noncombustible tobacco products, suggesting improved airway responsiveness and platelet function.

Few clinical studies have compared thromboxane synthesis in combustible versus noncombustible tobacco product users. For example, smokeless tobacco users, despite having urinary cotinine levels similar to that of smokers, have urinary levels of 2,3-d-TXB₂ (52) and TXB₂ (53) similar to those of nontobacco users. Our study examined the effect of short-term switching on AA metabolites associated with platelet activation across a spectrum of noncombustible tobacco products including Orbs, snus, Sticks, Strips, and VS Original and VS Menthol. Consistently, use of noncombustible tobacco, including vapor products, was associated with reduced platelet activation, as indicated by reduced urinary excretion of 2,3-d-TXB₂. Platelet activation was also diminished in smokers switching to either dual use or strict tobacco abstinence. It is unclear why urinary 2,3-d-TXB₂ levels were reduced in the abstinence cohort (Study II) and not in smokers abstaining from tobacco use in the CPDR study (Study I). One possible explanation is that smokers in the CPDR study had 2,3-d-TXB₂ levels approximately 50% lower than smokers in the abstinence cohort (Study II).

Platelet activation represents a well-established physiologic response elicited by exposure to combustible tobacco toxicants and is known to contribute to the etiology and progression of cardiovascular disease (54, 55). Cigarette smoke toxicants induce an inflammatory airway phenotype in which activated platelets serve as a major cellular source of TXA₂ (50). For example, acrolein, a toxicant found in cigarette smoke, potentiates platelet aggregation and TXA₂ synthesis in response to thrombin by increasing the availability of AA, a substrate for TXA₂ formation (56). TXA₂, in turn, acts on TXA₂ receptors to induce the second-phase response to cigarette smoke (57). Our finding that short-term product-switching yields reduced 2,3-d-TXB₂, the most abundant urinary metabolite derived from TXA₂, suggests that 2,3-d-TXB₂ may serve as a BoPH of platelet activation.

Some limitations of this BoPH study include: (i) the relatively small sample sizes in all the three studies; (ii) the lack of a proportional representation of Hispanics in the studies; (iii) the need for refinement of the classification model; and (iv) the need for consideration of the influence of diet on AA metabolism. Enrollment of the subjects was open to all those who met the inclusion/exclusion criteria and inadequate representation of Hispanics could be due to the location of where the studies were conducted. Although the sample sizes for each of the three clinical studies were relatively small, the two BoPH show consistent changes in the direction and the magnitude. The two-biomarker–based classification model described above was built based on limited size of biomarker data and further refinement using independent larger sample size of clinical datasets may be necessary. Because diet and genetic variation are potential determinants of AA metabolite levels, these factors also might require further consideration (58).

In conclusion, enhanced thromboxane and leukotriene biosynthesis are linked to platelet activation, chemotaxis, and airway hypersensitivity, which are early pathophysiologic events in cardiovascular and pulmonary disease (Supplementary Fig. S6). 2,3-d-TXB₂ and LTE₄, relatively stable metabolites of TXA₂ and LTA₄, respectively, are used as proxies for thromboxane and leukotriene synthesis. In this study, we demonstrated that urinary 2,3-d-TXB₂ and LTE₄ levels are lower following reduced toxicant exposure resulting from product switching. 2,3-d-TXB₂ and LTE₄ can discriminate between smokers and smokers switched to either abstinence or noncombustible products over a period of approximately 1 week. Moreover, these two metabolites reflect altered AA metabolism, which is linked to enhanced platelet activation, leukocyte recruitment, and inflammation. Because 2,3-d-TXB₂ and LTE₄ possess these key attributes, they represent potential BoPH designed to gauge the reversibility of toxicant exposure response in short-term tobacco product–switching studies.

P. Makena is a master scientist for and reports receiving a commercial research grant from RAI Services Company. G. Liu is a senior scientist for and reports receiving a commercial research grant from RAI Services Company. P. Chen is a master scientist for and reports receiving a commercial research grant from RAI Services Company. C.R. Yates is a consultant for RAI Services Company. G.L. Prasad is a director for and reports receiving a commercial research grant from RAI Services Company. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Makena, G.L. Prasad

Development of methodology: P. Makena

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Makena, G.L. Prasad

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Makena, G. Liu, P. Chen, G.L. Prasad

Writing, review, and/or revision of the manuscript: P. Makena, G. Liu, P. Chen, C.R. Yates, G.L. Prasad

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Makena, G.L. Prasad

Study supervision: P. Makena, G.L. Prasad

Other (editorial communications): P. Makena

The studies and the manuscript preparation were funded by RAI Services Company. RAI Services Company is a wholly owned subsidiary of Reynolds American Inc., which is a wholly owned subsidiary of British American Tobacco plc. The authors sincerely thank Herman Krebs (RAI Services Company) for sample management and ABF GmbH for bioanalysis of this biomarker discovery study. In addition, the authors would like to thank Nasrin Nouri, who was an intern of RAI Services Company, and Quynh Tran, a former employee of RAI Services Company.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.