Study design and population

This cross-sectional study was conducted from April to August 2024 across seven Vietnamese provinces representing different geographical regions: Lang Son (Northeastern), Hoa Binh (Northwestern), Ha Noi (Red River Delta), Nghe An (North Central), Khanh Hoa (South Central and Central Highlands), Ho Chi Minh City (Southeastern), and Can Tho (Mekong Delta).

A total of 1,077 adults aged 18–60 years were recruited. Eligible participants were healthy at the time of sampling, without chronic diseases requiring treatment (e.g., hypertension, diabetes, chronic obstructive pulmonary disease, metabolic disorders, or cancer). Exclusion criteria included pregnancy, performing strenuous physical activity, direct exposure to chemicals at work, or living near industrial pollution areas.

Study participants were selected via multi-stage sampling procedure as follow:

1. Stage 1. Selecting provinces: To ensure that the research sample represents the Vietnamese working-age population in different parts of Vietnam, we randomly selected one province from each of seven administrative and socio-ecological regions of the country. The selected provinces included Lang Son province (Northeast region), Phu Tho province (old Hoa Binh, Northwest region), Hanoi city (Red River Delta region), Nghe An province (North Central region), Khanh Hoa province (South Central Coast and Central Highlands region). Ho Chi Minh city (Southeast region), Can Tho city (Mekong Delta region).
2. Stage 2. Selecting commune/ward areas: In each province/city, we randomly selected one district/county. In each district/county, we intentionally selected one ward/commune that was not located near mining, metallurgy, pesticide production, large industrial zones, had no craft villages with the risk of exposure to toxic chemicals, and no cultivation of industrial crops. Seven ward/communes were invited to join the study.
3. Phase 3. Selecting subjects. Local health workers under the guidance of study staff prepared a list of subjects who met the study inclusion criteria. Subjects were stratified by sex (approximately 1:1 male-to-female ratio) and across four age groups (18–24, 25–34, 35–44, 45–60 years). From the above list, 20 subjects were randomly selected in each age group with equal sex ratio. Thus, 160 subjects (80 men and 80 women) were invited to join the study in each commune/ward. During the data collection process, if the subject refused to join or end their participation, another study subject would be randomly selected from the initial list. In total, 1077 healthy Vietnamese people in working age participated in the study.

Data/ Sample collection

Data were collected through structured, face-to-face interviews by trained staff at the time of urine sample collection. The structured questionnaire was developed specifically to gather information on participants’ characteristics. A pilot test was conducted with a small sample of participants to evaluate the comprehensibility, estimated completion time, and logical flow of the questions. Revisions were made based on pilot findings, including adjustments to wording, phrasing, and item sequencing. The finalised version of the questionnaire was used for formal data collection.

The study staff sought for approval from the Commune People’s Committee for the study implementation. Trained local healthcare workers served as interviewers for the study. Invitation letters were sent to participants in advance of scheduled sessions at the commune or ward health station. Upon arrival, participants were provided with information about the study’s aims and procedures. Written informed consent was obtained before participation. Five data collectors conducted face-to-face interviews, each lasting approximately 20 minutes. Completed questionnaires were collected immediately upon conclusion of each interview.

Urine samples were collected using standardised self-collection kits after interviews at study sites. Participants provided midstream urine samples to minimise contamination. Approximately 3–4 mL was collected for creatinine analysis, and 45 mL for cotinine analysis. Samples were transported at 4–8°C and stored at −20°C until analysis. Participant recruitment and data/sample collection were conducted concurrently from 21/05/2024 to 05/07/2024.

Measures

The questionnaire covered demographic information, lifestyle behaviours, including smoking and tobacco smoke exposure at home. Demographic data included age (18–24; 25–34; 35–44 and 45–60 years), sex (male or female), marital status (married and others), education level (elementary, junior high school, high school and college and above), occupation (homemakers, farmers, workers, officials and others professions), and monthly income (< 1 million, 1–5 million, 5–10 million, > 10 million VND). Lifestyle habits were assessed, including work-related travel, which was categorised into three levels: rarely, occasionally, and frequently. Alcohol consumption and physical exercise habits were recorded as binary responses (yes or no).

Exposure to tobacco smoke was recorded through three variables: Active smoking, Second-hand smoke exposure at home and Non-smoking exposure at home. Active smoking was defined as currently smoking either cigarettes or pipe tobacco. Second-hand smoke exposure at home was characterised by a non-smoking individual’s exposure to second-hand tobacco smoke resulting from the presence of smokers within the household. No smoking at home included individuals who reported no smoking and living in households free of smokers.

The full questionnaire is available in S1 File.

Laboratory analysis

Urine analysis was conducted at the Department of Medical Testing and Environmental Analysis, National Institute of Occupational and Environmental Health (NIOEH), accredited under ISO/IEC 17025:2017 including nicotine/cotinine technique. Cotinine concentrations were quantified using high-performance liquid chromatography (Agilent 1290 Infinity HPLC) coupled with tandem mass spectrometry (QTRAP 3200, SCIEX), following a validated method [25].

Sample preparation involved solid-phase extraction using HLB columns (60 mg/3 mL). Columns were preconditioned with methanol and deionised water, and 1 mL of urine was loaded. Analytes were eluted with 1mL of 10% formic acid in methanol and 10% formic acid (1:1). Certified reference materials (ClinChek® Urine Controls, Germany) ensured quality control. The limit of detection (LOD) was 0.03 µg/L for cotinine, with a mean recovery rate of 101.9%.

Certified reference materials (CRM), ClinChek® Urine Controls, Germany level I, were used for quality control the results of LC MS/MS analysis. Measured results and recoveries for cotinine in CRM ClinChek® Urine Controls, Germany were shown in the Table 1. CRM had a target value of 163 µg/L, the allowable value range was from 130 µg/L to 196 µg/L.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Measured results and recoveries for cotinine in Certified reference materials ClinChek® Urine Controls.

https://doi.org/10.1371/journal.pone.0338434.t001

In epidemiological research, spot urine sampling is frequently preferred due to its methodological advantages: it involves the collection of a single specimen, making it logistically simple and minimally disruptive to participants’ daily routines, while also reducing material and operational costs. Nonetheless, spot urine samples are susceptible to variability in analyte concentration resulting from fluctuations in urine dilution. Therefore, biomarker concentrations should be adjusted for urinary creatinine concentrations. The normalisation of analyte concentrations to urinary creatinine levels is a widely adopted approach to correct for variations in urine dilution, thereby enhancing the analytical accuracy and inter-sample comparability [26,27]. In this study, a cotinine-to-creatinine ratio (CCR) was used as a biomarker of tobacco smoke exposure.

Statistical analysis

Statistical analyses were conducted using SPSS version 20.0 (IBM Corp., USA). Cotinine concentrations below the limit of detection (LOD) were imputed as LOD divided by 5. Descriptive statistics included the number of observations (N) and corresponding percentages (%). Given that the distribution of the cotinine-creatinine ratio (CCR) was non-normal, the median and interquartile range (IQR) were reported, and group differences were assessed using the Mann-Whitney U test and Kruskal–Wallis test.

Cotinine concentrations normalized by urinary creatinine were evaluated for extreme values using multiple statistical approaches, including the interquartile range (IQR) rule, z-scores (> |3|), and the 99th percentile threshold. A biological plausibility check was also performed based on published reference ranges for urinary cotinine in smokers. Data points that were identified as extreme by at least two independent methods were classified as outliers. Using this combined approach, 32 outliers (2.97% of total samples) were flagged. These values were retained in the dataset with an indicator variable for sensitivity analyses and model robustness checks.

To identify the optimal CCR cut-off for distinguishing smokers from non-smokers, Receiver Operating Characteristic (ROC) analysis was employed, following the method introduced by Youden (1950) and widely adopted in subsequent research [28–30]. This approach identifies the cut-off point that maximises the difference between the true positive rate and false positive rate, aiming to classify individuals as accurately as possible. The ROC curve plots sensitivity against 1-specificity for all potential cut-off values of CCR.

A CCR cut-off point yielding an area under the ROC curve (AUC) between 0.8 and 1.0 indicates excellent discrimination of active smokers. An AUC between 0.7 and 0.8 reflects moderate accuracy, while values below 0.7 suggest poor discriminatory ability. An AUC of 0.5 or lower implies no meaningful distinction between smokers and non-smokers based on urinary CCR [30]. The distribution of tobacco smoking status based on the optimal CCR cut-off point across participant characteristics was evaluated using Chi-square tests, odd ratios and 95% confidence intervals (CI). Statistical significance was set at p < 0.05.

Ethics statement

The study protocol was approved by the Ethics Committee in Biomedical Research at the National Institute of Occupational and Environmental Health, Ministry of Health, Vietnam (Certificate No. 01/GCN-HDDD, issued March 29, 2024). All participants provided a written informed consent before enrolment. Personal information was kept confidential and accessed only by the main investigator for this study’s purpose.

Optimal cut-off point of CCR to classify active smokers and non-smokers

The CCR is a well-established biomarker for recent nicotine exposure [34–37]. In this study, the optimal cut-off value of 20.947 µg/g creatinine was determined using ROC analysis, which yielded an area under the curve (AUC) of 0.83, indicating good discriminative ability. The sensitivity (70.6%) and specificity (90.1%) at this cut-off demonstrate that the CCR can effectively differentiate smokers from non-smokers in the general adult population.

This cut-off value is consistent with findings from other population-based studies in Asia and Europe, which have proposed CCR thresholds ranging from 15 to 30 µg/g creatinine depending on demographic and environmental factors [37–41]. However, our estimated threshold (20.947 µg/g) is notably lower than the value reported by Yang et al. (2022), who analysed KNHANES VII data of 3,203 Korean participants aged ≥ 6 years and used a urine cotinine cut-off of 100 µg/g creatinine [42]. It is also lower than the range suggested in Kim’s literature review (2019), which reported cut-offs between 33 and 131 µg/g creatinine [43]. These variations across studies may reflect differences in race and ethnicity, smoking behaviour, environmental tobacco smoke exposure, and the implementation of smoke-free policies. In addition, the urinary cotinine cut-off levels were also varied with different measurement methods [44].

Nevertheless, our cut-off value represents a reliable and context-specific biomarker threshold for accurately classifying tobacco exposure in Vietnam, where underreporting of smoking remains a concern. Importantly, CCR offers a key advantage over self-reported data, as it provides an objective measure of exposure unaffected by social desirability bias or recall limitations—issues commonly encountered in tobacco research [14,41,45,46]. This objective measurement is particularly valuable for evaluating the effectiveness of smoke-free policies and for identifying populations at high risk of tobacco exposure.

Misclassification between self-reported exposure and optimal CCR

Despite the overall correlation between CCR and self-reported exposure, our findings highlight the discordance on smoking status, particularly among smokers. Based on the optimal CCR cut-off, 21.0% of participants were identified as active smokers, higher than the 18.3% who self-reported smoking. This finding is similar to that of previous studies where biomarker results were higher than self-reported results [37,38,41]. This discrepancy suggests an underreporting rate of approximately 2.7%, which is consistent with literature citing the social stigma and reporting bias in smoking behaviours, especially among women or in regions with strong anti-smoking norms [47]. On the other hand, these findings may suggest an excessive environmental exposure to tobacco smoke, usually SHS and THS in Vietnam, especially women who highly expose to tobacco smoke at home and at work [48]. In our study, 19.2% SHS and 3.5% participants from non-smoking households had the CCR level of active smokers. Biomarkers in general and urinary CCR are suitable tools to quantify the tobacco exposure level among SHS and THS which is impossible with self-reported methods.

False positives (CCR ≥ 20.947 µg/g creatinine among self-reported non-smokers) occurred in 9.9% of cases, potentially reflecting misclassified SHS exposure or recent but unreported smoking activity. Conversely, 29.4% of self-reported smokers were misclassified as non-smokers based on CCR < 20.947 µg/g creatinine, possibly due to recent abstinence or interindividual differences in cotinine metabolism [49].

These findings reinforce the importance of incorporating biochemical verification in population-based smoking surveillance. Overreliance on self-report alone may underestimate true exposure and undermine efforts to monitor tobacco control progress.

For instance, the U.S. National Health and Nutrition Examination Survey (NHANES) 2017–2020 reported detectable serum cotinine in approximately 51.2% of non-smoking adults, despite only 22.0% reporting SHS exposure [14]. Similarly, the Korean National Environmental Health Survey (KONEHS) 2015–2017 found that 94.1% of non-smoking adults had detectable urinary cotinine, with geometric mean concentrations of 2.1 μg/L in smoking homes and 1.3 μg/L in smoke-free homes [16]. These findings highlight the limitations of self-reported exposure assessments and emphasise the value of biomarker-based monitoring. Urinary cotinine, with a half-life of approximately 16 hours, serves as a reliable indicator of recent nicotine exposure. The high prevalence of detectable cotinine among non-smokers suggests ongoing involuntary exposure to environmental nicotine sources, including SHS and THS.

Distribution of smoking status based on the optimal CCR cut-off across groups of study participants

The smoking status which was defined by CCR significantly varied across groups of participants with different sociodemographic and behavioural characteristics. Men had 6.6 times higher odds of being smokers than women, consistent with regional trends and prior research in Southeast Asia [50]. Participants from Ho Chi Minh City and Khanh Hoa also exhibited significantly higher smoking rates compared to those from Hanoi, suggesting regional variations potentially influenced by urbanisation, lifestyle factors, or enforcement of smoke-free laws.

Education emerged as a protective factor, with participants who completed college or higher education significantly less likely to be smokers. This supports global evidence linking higher education to greater awareness of smoking risks and adherence to health-promoting behaviours [2,51,52]. Similarly, being married was associated with reduced smoking likelihood, possibly reflecting social and familial responsibilities that deter tobacco use.

Income level demonstrated a non-linear association with smoking. Individuals in the highest income group (>10 million VND/month) were more likely to smoke, contrasting with trends in high-income countries where smoking is more prevalent among lower-income populations [2,52]. Our findings are also inconsistent with the previous study in Vietnam, which reported higher smoking prevalence among the lowest income quintile group [53,54]. This may reflect the change in Vietnam’s economic context, where rising income can increase access to new tobacco products. Meanwhile, the widespread availability and variety of tobacco products, including affordable and easily accessible e-cigarettes, contribute to the increased prevalence of both active smoking and exposure to second-hand smoke (SHS). This is particularly concerning among young people [55], as e-cigarettes are often marketed with appealing flavours and designs, making them attractive to youth and potentially leading to nicotine addiction and the use of other tobacco products.

Occupational status also played a role, with officials and farmers less likely to smoke compared to freelancers and informal workers. The latter group may experience more stress, irregular routines, or lower exposure to workplace smoking bans, contributing to higher smoking rates [56].

Work-related travel and alcohol consumption were strong behavioural correlates of smoking. Participants who frequently travelled for work had nearly twice the odds of smoking, possibly due to stress, socialisation norms, or reduced oversight. Alcohol drinkers were three times more likely to be smokers—a relationship well-documented in addiction literature due to shared neurobiological pathways and co-use tendencies [57–59].

Demographic analyses reveal higher exposure levels among males, younger adults, and individuals with lower educational attainment, paralleling global evidence linking tobacco exposure to social determinants of health [60–62]. These patterns are consistent with findings from Latin American studies, where vulnerable populations exhibited disproportionate exposure in environments with weak policy enforcement [18]. These demographic disparities highlight the need for targeted public health interventions aimed at reducing nicotine exposure among high-risk groups.

Despite its strengths, including a nationally representative sample, stratified sampling, and standardised biomarker protocols, this study has limitations. First, tobacco exposure was assessed only through a single urinary sample, reflecting short-term exposure (2–4 days). Although urinary cotinine is a robust indicator of recent nicotine exposure, it does not fully capture cumulative or long-term THS exposure. Second, we could not collect the first-void morning urine samples which was ideal for assessing cotinine levels. Due to the design of our community-based study, urine samples were collected onsite only when participants visited the research site. This approach was necessary because we also analysed the samples for metals and other metabolites and therefore could not ensure proper cleanliness and preservation if sampling materials were distributed in advance. To address potential variability in urine dilution, we normalized cotinine results using urinary creatinine concentrations. Third, the absence of environmental measurements (e.g., air nicotine monitors or surface residue) limits our ability to triangulate personal exposure with environmental contamination. Future studies should integrate multi-route assessments, including dust and surface samples, to better quantify THS [9,10].