Lung cancer is the second most common cancer and the leading cause of cancer-related death in both men and women in the United States (U.S.) (1). It is well established that cigarette smoking is the primary risk factor for lung cancer. However, disease risk is not equal among individuals, as it is estimated that 11–24% of smokers will develop the malignancy over their lifetime (2). Along with nicotine, each cigarette delivers a complex mixture of tobacco-related carcinogens and toxicants. There are 80 established carcinogenic constituents in cigarette smoke, including several lung carcinogens such as 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), multiple polycyclic aromatic hydrocarbons (PAH), and volatiles such as 1,2-butadiene and acrolein (3). Inter-individual differences in the risk of smoking-related lung cancer may be, in part, attributable to the variation in harmful constituent exposure. Moreover, the investigation of the association of these biomarkers with lung cancer risk may improve our understanding of the mechanisms by which tobacco exposure causes lung cancer (4,5).

Biomarkers of smoking, such as metabolites of nicotine, and those of tobacco carcinogens and toxicants, have been shown to better reflect the internal dose of these harmful constituents and in some studies, their levels correlate with lung cancer risk. For example, prior epidemiological studies have shown that internal smoking dose, as determined by either cotinine (6–15) or total nicotine equivalents (TNE; the molar sum of nicotine and five or six metabolites in urine) (12,13,16), was associated with lung cancer risk, independent of self-reported smoking history. Additionally, studies have shown that 4-(methylnitrosamino)-1–3-pyridyl)-1-butanol (NNAL; a biomarker for NNK) (8–13) and phenanthrene tetraol (PheT; a biomarker for PAH) (9–11,17) are associated with lung cancer, even after adjusting for cotinine or TNE. However, these prior studies were conducted in single populations (e.g., only Whites or Chinese from Asia), and do not address the variation in smoking behaviors and lung cancer risk across populations. In addition, cotinine, along with self-reported smoking history does not fully account for individual variability in nicotine metabolism, which influences smoking dose and exposure to tobacco toxicants (12,18–21). Cytochrome P450 2A6 (CYP2A6) is the primary catalyst of nicotine metabolism and among current smokers in the Multiethnic Cohort (MEC) study, smoking behavior and dose were influenced by a smoker’s urinary nicotine metabolite ratio (total trans-3’-hydroxycotinine [3-HCOT]/cotinine) (22). This ratio is a phenotypic measure of CYP2A6 enzymatic activity and in MEC current smokers we previously reported that it was associated with lung cancer risk even after adjustment for self-reported smoking history and TNE (16). Acrolein, a well-known lung irritant, was recently reclassified by the International Agency for Research on Cancer (IARC) as probably carcinogenic to humans (23). However, the contribution of acrolein exposure to lung cancer risk in smokers has not been well studied. Therefore, a systematic examination of common biomarkers of tobacco toxicant exposure and metabolism is needed and may provide information on lung cancer risk beyond that of self-reported smoking history.

To better understand the individual and joint effects of smoking dose and specific smoking-related toxicant exposures on lung cancer risk, we evaluated the associations of a number of tobacco exposure biomarkers with the risk of smoking-related lung cancer among smokers in the Multiethnic Cohort (MEC) Study (n=140 incident lung cancer cases over an average of 13.4 years of follow-up). These biomarkers included NNAL, S-phenylmercapturic acid (SPMA, a biomarker of benzene uptake), 3-hydroxypropyl mercapturic acid (3-HPMA, a biomarker of acrolein uptake), PheT, 3-hydroxyphenanthrene (PheOH, a biomarker of PAH detoxification), the ratio of PheT/PheOH (proposed as a biomarker of metabolic activation of PAH), cadmium (Cd), and (Z)-7-[1R,2R,3R,5S)-3,5-dihydroxy-2-[(E,3S)-3-hydroxyoct-1-enyl]cyclopentyl]hept-5-enoic acid (8-iso-PGF_2α, a biomarker of oxidative stress).

Baseline characteristics of this cohort of MEC smokers (N=2,309) and the 140 incident lung cancer cases are presented in Table 1. Overall, the eligible population was comprised of men (46%) and women (54%) from five racial/ethnic groups: African Americans (16%), Native Hawaiians (14%), Whites (19%), Latinos (20%), and Japanese Americans (31%). Participants’ median age was 63 years, and their smoking history was a reported median of 23 pack-years. After an average of 13.4 years of follow-up from biospecimen collection, 140 incident lung cancer cases were identified. Cases were more likely to be male (55%) and there was a slightly greater population of African Americans (21%) and Native Hawaiians (22%) compared to the eligible population. Consistent with the analysis in the overall MEC study (35), age-standardized incidence rates (ASIRs) were highest in Native Hawaiians and African Americans, followed by Whites, Japanese Americans, and Latinos (Supplementary Table 2). Here, ASIRs were estimated only among this smoking population (N=2,309). Cases compared to the MEC eligible cohort were slightly leaner (median BMI:24.4 vs 25.7 kg/m²) and reported a greater number of pack-years (median:34.3 vs 23.3). The majority of lung cancer cases were diagnosed with adenocarcinoma (34%), followed by squamous cell carcinoma (29%), other/unspecified/non-specific (NOS; 17%), small-cell lung cancer (16%), and large-cell carcinoma (4%).

The adjusted geometric means and 95% CIs of all urinary biomarkers evaluated in this analysis are presented in Table 2 (minimally adjusted for just race/ethnicity in Supplementary Table 3). Compared with the overall subcohort of current smokers at the time of urine collection, incident lung cancer cases had higher geometric mean levels of all urinary biomarkers, except for PheOH, which was the same across groups. All urinary biomarkers were positively correlated with TNE levels (Supplementary Figure 2).

The associations of known risk factors with lung cancer risk were confirmed (Supplementary Table 4). The association between smoking-related toxicant and carcinogen biomarkers and lung cancer risk in this multiethnic population of current smokers is presented in Table 3. After adjusting for decade of birth, sex, race/ethnicity, BMI, and creatinine, a one-SD increase in log-pack-years was associated with a 90% increase in lung cancer risk in current smokers (Table 3, Model 1). After adjustment for birth, sex, race/ethnicity, BMI, and creatinine and pack-years of smoking, we found that both a one SD increase of log TNE and a one SD increase of log total 3-HCOT/cotinine ratio were associated with lung cancer risk (HR = 1.36, 95% CI: 1.00–1.84 and HR= 1.37, 95% CI:1.11–1.71, respectively; Table 3, Model 1). When TNE was further adjusted for total 3-HCOT/cotinine, the association was attenuated and no longer remained statistically significant (HR per SD increase in log-TNE = 1.22, 95% CI:0.91–1.64; Table 3, Model 2). Whereas the ratio of total 3-HCOT/cotinine remained significantly associated with lung cancer when adjusted for TNE (HR per SD increase in log-total 3-HCOT/cotinine = 1.33, 95% CI: 1.06–1.66; Table 3, Model 2).

Smoking-related toxicant levels of urinary 3-HPMA and Cd were individually associated with lung cancer risk after adjusting for decade of birth, sex, race/ethnicity, BMI, urinary creatinine, and self-reported pack-years (HR per SD increase in log-3-HPMA = 1.46, 95% CI:1.14–1.88 and HR per SD increase in log-Cd = 1.48, 95% CI:1.21–1.82; Table 3, Model 1). These associations remained even after adjustment for TNE (HR per SD increase in log-3-HPMA = 1.41, 95% CI:1.07–1.85 and HR per SD increase in log-Cd = 1.45, 95% CI:1.18–1.79). 3-HPMA association was driven primarily by those who smoked less intensely (TNE < median 32.4 nmol/mL HR=1.78, 95% CI:1.17–2.69 versus TNE ≥ median 32.4 nmol/mL HR=1.16, 95% CI: 0.80–1.70; Supplementary Table 5) and in a population who smoked less on average (Japanese Americans, n=29; HR = 1.96, 95% CI: 1.14–3.36; p for interaction race/ethnicity = 0.029; Table 4). The association for Cd was primarily driven among those who smoked more intensely (TNE < median 32.4 nmol/mL HR=1.42, 95% CI:1.07–1.89 versus TNE ≥ median 32.4 nmol/mL HR=1.44, 95% CI:1.01–2.04; Supplementary Table 5) and by the results in Whites (n=23; HR Cd = 1.91, 95% CI: 1.12–3.24; p for interaction race/ethnicity = 0.383).

We also explored the association by histological cell-type (Table 5) and found that the association between incident lung cancer risk and pack-years was strongest for adenocarcinoma (ADC; p<0.0001; Table 5). The association with TNE was strongest for risk of SCC (p=0.038). The association with 3-HPMA was strongest for risk of lung cancer of other/unspecified histologic cell-types (p=0.008). The association with urinary Cd was strongest for risk of ADC (p=0.001). These associations were not found heterogeneous across histologic cell-types. Associations of race/ethnicity and by histologic cell-type were not conducted due to small sample size (Supplementary Table 6).

This is the first study to examine the individual and joint effects of urinary biomarkers of multiple classes of smoking-related toxicants with the risk of smoking-related lung cancer in a multiethnic population. We demonstrated that urinary biomarkers of total 3-HCOT/cotinine, 3-HPMA (a metabolite of acrolein), and Cd (a biomarker of long-term Cd exposure) were significantly associated with lung cancer risk, independent of self-reported smoking history (pack-years) and internal smoking dose (TNE).

CYP2A6 catalyzed 5’-oxidation, the primary pathway of nicotine metabolism in most smokers produces cotinine (21), the most commonly used biomarker of tobacco exposure (21,36). However, cotinine levels are influenced by individual variability in nicotine metabolism, which is known to influence exposure to tobacco toxicants (12,18–21,37). TNE, which includes metabolites from three pathways of nicotine metabolism accounts for ~85% of the nicotine dose and better reflects nicotine uptake, internal smoking dose, and lung cancer risk in some populations (4,21). Since CYP2A6 also catalyzes the oxidation of cotinine to 3-HCOT, the ratio of 3-HCOT/cotinine is a phenotypic measure of CYP2A6 enzymatic activity. Variation in CYP2A6 activity can influence smoking intensity (30,37,38) and the ability to quit smoking over time (39) and therefore the urinary ratio of 3HCOT/cotinine may reflect longer-term smoking behaviors that are not captured by the short-term biomarker of dose (e.g., TNE).

Previously, we reported, in the same population studied here that a one log unit increase in the ratio of 3-HCOT/cotinine was associated with a 52% increase in lung cancer risk (n=92 cases), independent of CPD, self-reported smoking duration, and TNE (16). In the current analysis with an additional 48 cases, the ratio of total 3-HCOT/cotinine was also significantly associated with lung cancer risk after adjustment for similar covariates (this analysis adjusted for the measure pack-years instead of CPD and self-reported smoking duration). These data support our hypothesis that this biomarker provides additional information regarding smoking history that may not be captured by self-reported data (16). In the Singapore Chinese Health Study, a positive association between total 3-HCOT/cotinine and lung cancer was observed, however, after adjustment for TNE, the association was attenuated (13). It is not surprising that in different populations with different genetics of nicotine metabolism and tobacco use, the ability of TNE and/or the ratio of total 3-HCOT/cotinine to reflect long term smoking behavior varies.

Acrolein, an α,β-unsaturated aldehyde, is a ubiquitous environmental and dietary constituent and is produced endogenously as a product of lipid peroxidation, amino acid metabolism, and polyamine metabolism (40). Among current smokers, active tobacco smoking is the major source of exposure to acrolein (23,41). One recent analysis reported a range of acrolein levels from 30.8–82.6 μg per unit in the mainstream smoke (under International Organization for Standardization [ISO] conditions) of 35 commercial cigarette tobacco products sold in the U.S. (42). Acrolein is a known lung irritant that produces inflammation and various other effects involved in carcinogenesis. As such, in 2020, IARC reclassified acrolein into Group 2A as “probably carcinogenic to humans” (23). Aligned with this reclassification, we found that 3-HPMA (the major metabolite of acrolein) was associated with lung cancer risk in our MEC smoker population after adjusting for lung cancer risk factors, the short-term biomarker of dose (TNE), and common urinary biomarkers of smoking. This association was primarily driven by those that smoked less intensely and populations who smoked less on average (e.g., Japanese Americans and Latinos). The Shanghai Cohort Study of male current smokers (n=343 lung cancer cases) (43) and never-smokers only (n=82 lung cancer cases) (17) and the Golestan Cohort Study in Iran of male exclusive cigarette smokers (n=31 lung cancer cases) (44) found an association with acrolein (measured by 3-HPMA) and lung cancer risk. However, these studies found that the association was no longer statistically significant after adjusting for total cotinine or smoking history, respectively, suggesting that the observed effect between 3-HPMA and lung cancer risk partly reflected an added measure of smoking dose.

Cadmium is a known human carcinogen (IARC group 1), and the primary source of exposure in most cases (e.g., leaving aside occupational exposure) is cigarette smoke (45,46). While toxicological studies in rodents and occupational studies in humans have shown that Cd is a respiratory toxicant associated with lung cancer, occupational studies cannot be extrapolated to the general population, and few studies have properly adjusted for smoking in their risk estimates (45,47–49). Three non-occupationally based prospective cohort studies in the U.S. (NHANES III and Strong Heart Study) and Belgium (CadmiBel Study) have shown a positive association between urinary Cd levels, a biomarker of long-term Cd exposure, and lung cancer risk after adjusting for smoking status and/or pack-years (50–52). In two of these studies, after removing current smokers at baseline (Strong Heart Study; 52) or including only never-smokers (U.S. NHANES III) (51) in the analysis, the association remained positive but weaker. In the current study, we further demonstrated a positive association between Cd and lung cancer incidence, independent of measures of self-reported smoking history. Interestingly, when we explored associations by histological cell-type, we demonstrated a strong positive association between urinary Cd and adenocarcinoma in our population, which is in agreement with previous evidence that has shown chronic Cd inhalation in rodents causes pulmonary adenocarcinomas (53). These results, in combination with those of other studies, suggest that urinary Cd may provide additional information regarding smoking-related lung cancer risk.

Prior studies have also detected associations with NNAL (8–10), PheT (10,11,17), PheOH (17), SPMA (43), and 8-iso-PGF_2α (54,55) with lung cancer risk, even after accounting for self-reported smoking and/or smoking dose (cotinine and/or TNE). In this current study, we did not detect an association between these biomarkers and the risk of lung cancer with adjustment for pack-years and TNE. These reported differences across studies may be a result of differences in sample size (MEC with only 140 lung cancer cases), smoking behaviors, and study population (e.g., the Shanghai population smoked mainly Chinese cigarettes in the 1980s versus the multiethnic U.S. population smoked mainly domestic cigarettes in the 2000s). While all lung cancer histologic cell-types are attributed to tobacco smoking, the cigarette quantity and composition have been shown to influence lung cancer subtypes. For instance, cigarette smoking dose was associated with higher relative risks for squamous cell carcinoma and small cell lung cancer than adenocarcinoma (56). Additionally, increased filtration ventilation of cigarettes has been suggested as a potential contributor to the relative rise of adenocarcinoma (57). Our exploratory analyses stratified by race/ethnicity and by histologic cell-type further supports the possibility that variation in exposure to smoke toxicants may contribute to etiologic differences in lung cancer risk.

The main strength of this study is the use of a well-characterized multiethnic population of current smokers with a range of smoking intensity and our ability to evaluate the effects of nine urinary biomarkers of smoking-related toxicants with lung cancer risk independent of self-reported smoking history and urinary TNE (a biomarker for internal smoking dose). However, our study has some limitations. First, this study included a modest number of lung cancer cases, limiting the power of our racial/ethnic and histological cell-type-specific analyses. Second, the biomarker levels were only measured from a single urine collection; therefore, we did not account for variations in smoking behavior and exposure to tobacco carcinogens over time. However, adult smoking behavior is relatively stable over time (58). Third, we did not have information on quitting attempts or quitting during follow-up or information on cigarette brands, which could contribute to differences in biomarker levels of smoking-related toxicants.

In conclusion, our findings suggest that urinary total 3-HCOT/cotinine, 3-HPMA, and Cd levels may provide additional information on smoking-related lung cancer risk that is not captured by self-reported smoking history or the short-term biomarker of internal smoking dose (TNE). Our findings also suggest that these biomarkers may provide distinct information for lung cancer risk prediction by population or histological cell-type. Replication in large prospective studies is warranted.