The prospective cohort observational design of data collection was conducted between September 15, 2021, and September 15, 2022. We collected data from higher education students in colleges and universities in lower northern Thailand, including Phitsanulok, Phichit, Kamphaeng Phet, Uthai Thani, Chainat, Nakhon Sawan, Phetchabun, Tak, and Sukhotai. All individual participants included in the study gave their informed consent. The study protocol was approved by the Institutional Review Board (IRB) of Naresuan University (IRB no., P3-0110/2564).

A self-reporting online questionnaire was distributed among college and university students. Of 1,373 students, 133 current cigarette users who met the following criteria were included: (1) the age of at least 18 years old; (2) having personal devices such as mobile phones or computers, which they can use to response to the online questionnaires; and (3) responding “yes” to the two specific questions, which are “Have you ever used a cigarette?” and “Have you used a cigarette during the last 30 days?” (Standard National Survey on Drug Use and Health Current Cigarettes Using Variable Definition, NSDUH-S) [18]. If students answered “yes” to these questions, they will be defined as “current cigarette users” and (4) consenting to participate in the study. The 133 current cigarette users, including 58 e-cigarette users (43.6%), 33 combined users (24.8%), and 42 conventional cigarette users (31.6%) (Figure 1), received the self-reporting online questionnaires sent to them at baseline, 90, 180, 270, and 360 days to collect information on exposure, outcome, and confounders.

We used exact probability tests and a one-way analysis of variance to test for differences in proportions and means for categorical and continuous variables, respectively.

In our initial analysis, we treated both cigarette-use types and all confounders as “time-independent.” A generalized estimating equation (GEE) was used in this analysis. In the univariable analysis, we evaluated the effect of types of cigarette use, considered as “time-independent,” on FTND scores. Subsequently, in the multivariable analysis, we adjusted for all confounders, which were also treated as “time-independent.” We reported differences in mean FTND scores across five time points between index groups and the reference group as adjusted β-coefficients with 95% confidence intervals (CI). We also considered the magnitude of the effect of smoking types on the change in FTND scores depending on the level of time, which was called “the interaction between smoking types and time (group*time effect),” using GEE.

We used the marginal structural model (MSM), which is a form of GEE that considers the correlation among repeated measuring variables, to analyze the time-dependent smoking exposure variables. Because we considered time-dependent confounders in the analysis, we employed the inverse probability of weighting (IPW) of GEE [22], which could average the weight of time-dependent confounders at each time point to adjust for time-dependent confounders. All time-independent confounders were included in the analyses. The results were also reported as differences in mean FTND scores (adjusted β-coefficient) and a 95% CI.

The data were arranged into 3-month intervals that corresponded to planned appointments (at 90, 180, 270, and 360 days). Types of cigarette-use and FTND scores were time-dependent variables collected simultaneously at each time point. As a result, the loss-to-follow-up individuals would represent the entire absence of cigarette-use types and FTND scores. The loss-to-follow-up rates were 2.3%, 3.1%, 4.0%, and 6.6% at each time point, respectively, as shown in Figure 1. As the loss-to-follow-up rate remained below 10% at each interval, its impact on our overall findings was deemed as negligible. Moreover, we found that the participants’ characteristics did not differ between the missing and complete cases in the analysis, as shown in Supplement 1. Consequently, this does not necessitate the implementation of additional data.

All analyses were done with R Studio ver. 4.2.1 (RStudio, Boston, MA, USA).

In univariable analysis with GEE approach, including the interaction term (group*time effect), the difference in mean FTND scores when compared to the reference group was lower by 0.66 points in e-cigarette users (95% CI, -1.57 to 0.24) and by 1.02 points in combined cigarette users (95% CI, -2.31 to 0.26). However, the difference was not statistically significant. Furthermore, there was no interaction between types of cigarette-using groups and time (β-coefficient, -0.03; 95% CI, -0.21 to 0.15).

In our initial multivariable analysis, we considered the types of cigarette-use as time-independent exposures. Age at cigarette-use initiation, sex, income, alcohol consumption, and motivation to quit smoking were used as time-independent confounders. After adjustment for time-independent confounders in model 2, the result shows that the mean FTND scores of e-cigarette users were reduced by 0.82 points compared to the reference group. However, the reduction was not statistically significant (95% CI, -1.63 to 0.02). Similarly, the mean FTND scores were reduced by 1.40 points in combined cigarette users when compared to the reference group, but this reduction was statistically significant (95% CI, -2.32 to -0.49). Similar findings were observed in models 1 and 2.

After considering time-dependent exposure and confounders identified by the DAG (Figure 2), we used MSM with the GEE approach, adjusting for age at initiation of cigarette-use, sex (time-independent confounders), income, alcohol consumption, and motivation to quit smoking (time-dependent confounders) through IPW to balance confounders across each type of cigarette-use. The results indicate that e-cigarette users exhibited mean FTND scores that were 0.20 points lower (95% CI, -0.64 to 0.25) compared to the reference group, while combined users showed mean FTND scores that were lower by 0.47 points (95% CI, -0.96 to 0.01) compared to the reference, as evidenced in Table 3. However, the decrease was not statistically significant.

The main effects observed in our study, wherein e-cigarette-use and combined use resulted in slightly decreased FTND scores compared with conventional cigarette-use, are consistent with the findings of previous studies. For instance, e-cigarette users had slightly lower odds of nicotine dependence than conventional cigarette users (adjusted OR, 0.96; 95% CI, 0.80 to 1.15) [23]. Since nicotine dependence levels served as a monitoring tool in the smoking cessation process, previous studies have shown that e-cigarette users are more likely to quit smoking compared to conventional cigarette users (adjusted OR, 1.56; 95% CI, 1.12 to 2.18) [8]. Another study demonstrated that e-cigarette-use increased the odds of smoking cessation in comparison with conventional cigarette-use (adjusted OR, 1.63; 95% CI, 1.17 to 2.28) [7].

However, in the present study, the change in FTND scores over a 1-year period was not significant for e-cigarette users or combined users compared to conventional cigarette users. Some differences between our present study and previous studies are as follows: (1) The outcome was measured at only one time point [6-9,11]; however, nicotine dependence levels or smoking cessation outcomes may have changed during the study period. For example, nicotine dependence levels increased at the beginning of the period but may decline by the end. At baseline, a smoker could quit smoking; however, the smoker may begin smoking again at later periods. (2) Most studies assessed each type of cigarette-use only once at baseline [6-8]. Using this measurement of exposure, it was assumed that the types of cigarette-use would not change over the monitoring period (time-independent exposure). However, there might be a shift in cigarette-using patterns as e-cigarette users may switch to conventional cigarette-use later. (3) Most studies considered time-independent confounders [6-9,11], such as measuring motivation to quit smoking only at baseline; however, motivation to quit smoking may change over a 1-year period (e.g., a cigarette user who reports low motivation to quit at baseline may report high motivation to quit at 1 year). (4) Combined users were not considered (those who smoked both e-cigarettes and conventional cigarettes) [6-9,11]; thus, the effect of e-cigarette-use on the outcomes might be invalid. Therefore, all cigarette users were included in this study.

In the present study, we conducted an analysis that considered repeated measurements of the outcome over a 1-year period. We treated the types of cigarette-use and confounders as time-dependent variables, which allowed us to address the methodological concerns of previous studies. Furthermore, our analysis included a combined group of cigarette users. We observed a decline in FTND scores compared with conventional cigarette users.

The stable nicotine dependence levels over a year in e-cigarette and combined cigarette users compared to those using conventional cigarettes may be attributed to the oxidizing free-based nicotine produced by e-cigarette devices [24]. This form of nicotine, which is highly addictive and readily absorbed into the bloodstream through inhalation, can sustain nicotine dependence by activating nicotine receptors. Additionally, individuals who use e-cigarettes and combined cigarettes often find themselves in settings conducive to nicotine inhalation, such as being around other smokers [25], thereby maintaining consistent levels of nicotine dependence.

Our study had several strengths. First, the study was conducted on a cohort of higher education students in nine provinces representative of smoking among young adults in Thailand’s lower northern region. Second, we prospectively collected new data from cohorts, especially for repeated measurements of FTND scores, and considered exposure and confounders as time-dependent variables at five time points. Therefore, our study successfully addresses the methodological issues encountered in previous studies. Additionally, we used a regression method that accounted for time-dependent exposure and confounders to analyze the change in mean FTND scores over a 1-year period.

Our study has some limitations. Using a self-reporting questionnaire may introduce misclassification bias in exposure and outcome. For example, e-cigarette users may report themselves as conventional cigarette users. Thus, the effect of e-cigarettes on FTND scores may be attenuated. Nevertheless, the online questionnaire in our present study employed skip logic and branching logic techniques that minimized bias by modifying questions in accordance with past responses. Additionally, the actual behavior of e-cigarette users is not well understood. Different types or brands of e-cigarettes have different nicotine concentrations, which can also affect the measurement of nicotine dependence level under study. This challenge stems from products that do not have labels indicating their nicotine content, making precise measurements tricky. Moreover, legal issues in Thailand have made it challenging to assess and research the nicotine content in e-cigarettes, as well as to identify the type or brand of e-cigarettes used. Legal restrictions on e-cigarettes in Thailand further complicate this study.

Owing to the non-significant decline in nicotine dependence levels among e-cigarette users and combined users compared to conventional cigarette users and the unknown long-term health effects, healthcare practitioners in smoking cessation clinics should refrain from using e-cigarettes as the initial approach to reducing nicotine dependence in current cigarette users. Instead, a range of methods, including medical provider counseling [26], nicotine replacement therapy [27] and non-nicotine drugs [28] should be considered. More research is required to determine whether e-cigarettes assist smokers in quitting smoking. However, there are currently no established methods for measuring smoking cessation among e-cigarette and combined-cigarette users. A standard method to define smoking cessation in e-cigarette and combined-cigarette users requires further research.

E-cigarette use or the combined use of e-cigarettes and conventional cigarettes does not effectively minimize nicotine dependence in young adults who currently use cigarettes. Advocating these methods to curtail nicotine dependence in this population is not recommended. As there is no standard method for determining smoking cessation among e-cigarette and combined-cigarette users, further research is required to evaluate how e-cigarettes affect smoking cessation.