E-cigarette Use for Smoking Reduction and Cessation in a Four-year Follow-up Study Among Young Swiss Men: Some may Benet, but they are Few.

Randomised controlled trials have shown some benets to using e-cigarettes (ECs) to facilitate smoking reduction and cessation, but real-world observational studies have rarely conrmed this. The present study looked at EC use and smoking during a four-year longitudinal study of young men. It compares general population ndings with a subgroup of individuals with a good prognosis for quitting smoking. The smoking habits of 5353 young men at t1 (average 21.3 years old) were dened as either never-smoker, ex-smoker, initiates, relapsed smoker or persistent smoker. At follow-up (t2), smoking status was analysed using logistic regression, differences in the number of cigarettes smoked by persistent t1/t2 smokers were analysed using mixed linear models and the number of quit attempts was analysed using negative binomial models. a cohort of young men aged around 21.5 years old at baseline and continuing for 4 years of follow-up. It compared never-smokers, former smokers and persistent smokers in that population as a whole. This paper presents sensitivity analyses of the EC use modes assumed to have the most advantageous effects on smoking cessation (e.g. daily use, use of newer generations of ECs, use of nicotine liquids, and being motivated to reduce or cease CC smoking). The study was the continuation of an earlier one (49) which measured vaping at the baseline only. were calculated to test associations between vaping and cigarette-smoking status at t2. Vaping was, if possible, analysed according to its occurrence at t1, t2 or both times. If sample sizes were too small, we grouped vaping at t1 with vapers using EC at both time points to distinguish users who used EC cigarettes already at t1 from those who used ECs only at t2. To estimate changes in the number of cigarettes smoked by persistent smokers at t1 and t2, we calculated linear mixed-models which considered correlations within subjects, vaping as the between-subject factor, time of assessment (t2 vs t1) as the within-subject factor, and the interaction between vaping and time of assessment. The interactions tested the differences in cigarette-smoking changes between never-vapers and groups of vapers (dened by their vaping status at t1 and t2). Associations between vaping and the number of attempts to quit by t2 among smokers were tested using negative binomial regression models for counts of attempts to quit. cigarettes smoked (time). both time both time points (-4.39 signicant for the relative increase occasional smokers at t1 who vaped at t1 and t2: more cigarettes weekly than occasional cigarette smokers at t1 who vaped. had started smoking before their EC use, and we were able to monitor for vapers at t1 to t2—and similarly for smokers and non-smokers with regards to simultaneous or subsequent EC use. We were also able to say something about dual users at t1 and how they compared when they did or did not continue with EC use at t2. The present study overcame some of the problems inherent in other longitudinal studies: we were able to adjust for nicotine dependence and we followed-up with people who had quit smoking, not only treatment failures (13).


Introduction
Nicotine use creates dependence; however, it is unlikely to be directly responsible for much health harm. Most harm stems from nicotine delivery through smoking (1)(2)(3). Although smoking is decreasing in many-though not all-countries worldwide (4), its decline is seen as too slow (e.g., 1). This has led some scientists to ask whether harm reduction approaches may help to accelerate decreased smoking (3,5). Ecigarettes, e-hookas or pod mods may be used as harm-reducing electronic nicotine delivery systems (ENDS): their use is not harmless but is considerably less harmful than smoking (3,(6)(7)(8)(9)(10). We will use the term e-cigarette (EC) as a generic term. Arguments against ECs include that dual use may lead to sustained smoking among smokers who would otherwise have quit, that EC use may re-normalise smoking-like behaviour (inhaling) in public, and nally that ECs may divert smokers who are motivated to quit away from evidence-based smoking cessation treatments (6). ECs may also seem attractive to young people and lead to nicotine use which would not have occurred otherwise and subsequently to smoking (gateway effect : 11). A recent review of the evidence (12) identi ed seven systematic meta-analytical reviews, of which two found a positive effect (13,14), four were inconclusive (15)(16)(17)(18) and one found a negative effect (19). The present paper argues that the mixed ndings on the e cacy or effectiveness of using ECs for reducing or ceasing smoking may be related to the answers to our two research questions. Firstly, is there an overall general population, public health effect? Secondly, is there an effect among a subpopulation of heavy smokers who are highly motivated to quit smoking? The present longitudinal study investigated vaping and its associations with smoking among young men aged around 21.3 years old at baseline, with a follow-up around four years later, and it distinguished between a general "real-world" population perspective and a group of smokers motivated to quit.
One reason for the different conclusions of the meta-analytical reviews may have been the types of studies included. Whereas reviews suggesting bene cial effects were predominantly based on randomised control trials (RCTs) (3), the sole review suggesting negative effects (19) included numerous observational studies, and even cross-sectional studies, which could not separate cause and effect. The review by Villanti et al. (20) set stricter criteria for studies on using ECs for reducing or quitting smoking. Only four studies tted those criteria (21)(22)(23)(24). All four were randomized controlled trials (RCT) and found, at best, small, but usually non-signi cant favourable effects for using nicotine ECs.
Commentaries on the paper by Villanti et al. (20) criticised its focus on RCTs and its assessment limited to EC users who were only motivated by smoking cessation (25,26). To evaluate the potential impact of promoting ECs as a public health approach, there is a need to investigate how ECs are used in real-world clinical settings or, even more signi cantly, in the general population. Strict inclusion criteria limiting investigations to smokers who desperately want to quit (and excluding others) do not re ect real-world EC use in the general population. Many vapers do so for other reasons, e.g. costs, the ability to use it where tobacco is banned or to maintain a similar habit as part of daily routines.
Similarly, strict inclusion criteria as regards certain parameters of EC use, such as frequency of use, type, dose or duration, do not re ect realworld use either (26). Daily vaping and the use of more recent innovations in EC models (e.g. tank, mod and pod systems), which provide more effective nicotine delivery and therefore are su ciently appealing to smokers, have shown promising effects in RCTs (for reviews see 3, 20, 27-29) but do not necessarily correspond to real-world use at the general population level (see 30).
Although RCTs are commonly regarded as the strongest epidemiological designs, they do have inherent drawbacks. Simply participating in an experiment and having greater interaction with a therapist may increase motivation for the treatment, belief that the treatment is important and the behaviour typical of good subjects (31)(32)(33). Participants' informed consent may be perceived as binding contracts which they do not want to break (34), and the inclusion and exclusion criteria of RCTs commonly lead to highly selective groups of participants. This does not re ect the groups of patients found in standard clinical settings (35) and even less so in the general population.
Longitudinal studies in the general population are particularly important when it comes to regulating ECs (e.g. maximum nicotine concentrations, taxes). Even critics of ECs generally admit that their use can result in decreased smoking and cessation for some smokers.
However, whether the overall effect of EC use is harmful or bene cial depends on its predominant type of use in the general population. Those who regard EC use as more harmful than bene cial promote strong regulation (e.g. 11,36,37). Those who see its potentially bene cial effects, through decreased smoking and cessation, promote less regulation than for conventional cigarettes (CCs) (2,12,29).
Two of the recent systematic reviews (14,15), deemed to be of high quality (38), included three general population studies. Brose et al. (39) found that after one year, daily EC users were more likely than non-users to have attempted to stop smoking (but not to have stopped) and to have reduced their number of cigarettes smoked by 50%. No effects were found for occasional EC users, and only about one fourth of all EC users were daily users. Al-Delaimy et al. (40) found that, after one year, ever-EC users had made a non-signi cant greater number of attempts to quit, but they were signi cantly less likely to quit and less likely to have decreased cigarette use. Manzoli et al. (41) compared formerlysmoking EC users with tobacco smokers and dual users, and they found that EC users were signi cantly more likely to have persisted in their tobacco abstinence; however, dual users were not more likely to become tobacco abstainers, and they also did not reduce tobacco use.
There are recent longitudinal studies on the general population. Many looked at the potential for ECs to introduce young people to smoking (42)(43)(44), con rming reviews that ECs expanded the nicotine market by attracting young people who may later use CCs (11,45). A general population study of adults (18+) in the USA by Benharmina et al. (38) showed that ECs may increase the likelihood of tobacco abstinence among those attempting to quit. However, the vast majority relapsed to cigarette smoking, and those who did continued to use ECs in addition to CCs with no reduction of CC use. Findings from the same study (46) showed that at least high frequency of EC use at follow-up (but no use at baseline) may be associated with a reduction in CCs smoked. A general population study in France examined daily smokers and former daily smokers (47). After two years, regular (daily) vapers had reduced their daily CC use by about 1.7 cigarettes per day more than non-vapers. They also had a higher likelihood of quitting smoking. However, former smokers were more likely to relapse to CC smoking if they were vapers.
Daily users, but not occasional users, were also shown to reach higher abstinence compared with non-vapers (48).
The present study examined EC use and its associations with smoking cessation, attempts to quit and CC use among a cohort of young men aged around 21.5 years old at baseline and continuing for 4 years of follow-up. It compared never-smokers, former smokers and persistent smokers in that population as a whole. This paper presents sensitivity analyses of the EC use modes assumed to have the most advantageous effects on smoking cessation (e.g. daily use, use of newer generations of ECs, use of nicotine liquids, and being motivated to reduce or cease CC smoking). The study was the continuation of an earlier one (49) which measured vaping at the baseline only.

Sample
Data come from the rst three waves of the Cohort Study on Substance Use Risk Factors (C-SURF). Switzerland has a mandatory procedure for young men around 19 years old to determine their eligibility for military or civilian service. As men have to be at least 18 years old to be called for the procedure by the army, the youngest person in the sample was 18 years old. Conscripts were invited to participate in C-SURF during those procedures, but assessments were done independently of the army, at home, via the internet or on paper questionnaires. C-SURF is a larger study on substance use and behavioural addictions, not only exclusively on smoking and vaping. Questionnaires and other material is provided online at www.c-surf.ch. English versions of the questionnaires are also provided as supplementary les 1 to 3. C-SURF was approved by the Human Research Ethics Committee of the Canton Vaud (C-SURF: Protocol No. 15/07).
Of the 7556 conscripts who provided written informed consent to participate in the study, 5987 (79.2%) completed the baseline assessment (t0) from September 2010-March 2012. Of these, 5479 responded (91.5% response rate) to the rst follow-up assessment (t1) from March 2012-January 2014, and of these, 4981 responded (90.9% response rate) to the second follow-up assessment (t2) from April 2016-March 2018. We also invited consenting but baseline-non-respondent conscripts to participate in the t1 follow-up, and 391 completed both the t1 and t2 follow-ups and were thus included in the present study. These individuals had a known smoking status at t1 and t2, and logical imputation was used to impute their smoking status at baseline. Smokers and non-smokers at both t1 and t2 were also assumed, respectively, to have been smokers and non-smokers at t0. After eliminating 19 people with missing values on key variables, the nal sample size consisted of 5353 participants.

CC smoking
At all three time points, participants were asked whether they had smoked in the past 12 months and, if so, how often on a 6-point scale from "once a month or less" to "daily". All those smoking less than daily were considered "occasional smokers". Past-12-months smokers were asked about the number of cigarettes smoked on days when they did smoke, and their number of cigarettes per week was calculated using quantity and frequency. Additionally, at t0, participants were asked whether they had smoked at least 50 cigarettes in their life. Five groups were constructed according to their "CC status t0-t1": "never-smokers", i.e. never-smokers at t0 and t1; "ex-smokers", i.e. smokers at baseline but non-smokers at t1; "initiates", i.e. never-smokers at t0 but smokers at t1; "relapsed-smokers", i.e. ex-smokers at t0 but smokers at t1; and "persistent smokers", i.e. smokers at t0 and t1. Participants were followed-up as parts of these ve groups, and depending on sample size, some analyses required regrouping into smokers and non-smokers.

Outcomes
The primary outcomes considered were a) smoking cessation and b) a change in the number of cigarettes smoked by persistent smokers. The number of attempts to quit in the past 12 months was used as a secondary outcome. Possible responses were zero, one, two, three, and four or more attempts. Attempts to quit were de ned as seriously trying to stop smoking and not smoking for several consecutive days.

Vaping
Questions about vaping were only asked at t1 and t2. Frequency of EC use was measured on the same 6-point scale as CC smoking, but only for those participants who had reported vaping in the past 12 months. As the number of vapers was relatively small at t1, "almost daily vaping" was de ned as "5-6 days a week" or daily. For convenience, we named participants not vaping at t1 and t2 as "never-vapers". It should be noted that they may have used ECs before t1, although this was rather unlikely, or they may have used after t1 assessment but having stopped again at t2.

Dependence
Higher nicotine dependence has been positively associated with the use of smoking cessation aids and negatively associated with successfully stopping, but many longitudinal studies on ECs have been criticised for not accounting for levels of nicotine dependence (12,13,38,50). Nicotine dependence at t1 was assessed using the Fagerström Test for Nicotine Dependence (51) on a continuous scale scored from 0-10.

Covariates
Covariates were age, language (French, German), and the highest completed level of education at t1, which consisted of three categories: primary schooling (9 years); secondary vocational training (> 9-12 years); and post-secondary schooling (13 + years).

Statistical analysis
For descriptive purposes, we used percentages for ordinal and nominal variables and means with standard errors (SE) for continuous and count variables. Logistic regression models, strati ed by t1 cigarette-smoking status (never-smokers, ex-smokers, initiates, relapsed-smokers, persistent smokers), were calculated to test associations between vaping and cigarette-smoking status at t2. Vaping was, if possible, analysed according to its occurrence at t1, t2 or both times. If sample sizes were too small, we grouped vaping at t1 with vapers using EC at both time points to distinguish users who used EC cigarettes already at t1 from those who used ECs only at t2. To estimate changes in the number of cigarettes smoked by persistent smokers at t1 and t2, we calculated linear mixed-models which considered correlations within subjects, vaping as the between-subject factor, time of assessment (t2 vs t1) as the within-subject factor, and the interaction between vaping and time of assessment. The interactions tested the differences in cigarette-smoking changes between never-vapers and groups of vapers (de ned by their vaping status at t1 and t2). Associations between vaping and the number of attempts to quit by t2 among smokers were tested using negative binomial regression models for counts of attempts to quit.
Subgroup analysis of smokers with good prognostic values for quitting As a subgroup analysis, we looked at daily smokers at t1 who might be assumed to have a higher likelihood/intention/motivation to stop smoking by using ECs than occasional smokers, who were less addicted and might use ECs for other reasons (e.g. curiosity). We also examined almost-daily vaping (at least 5 days/week) as vaping intensity has been shown to have a more bene cial impact on reducing or ceasing smoking. Following the suggestion by Villanti et al. (20), we tried to de ne a group of vapers who might have a good prognosis for reducing or ceasing smoking. Vapers in this group had to be using at least a second-generation EC (with rechargeable tanks or pod mods) with liquid nicotine and to claim at least one of the following motives: a) use it in the course of a quit attempt, or b) use to stop smoking or not to relapse to smoking. This information was asked at t2 only.

Results
Smoking rates between t1 and t2 decreased from 47.4-41.2%, particularly occasional smoking. At the same time, the proportion of vapers almost doubled, from 5.2-9.8% (Table 1). Although the proportion of almost-daily vapers increased from 11.8-20.9%, the majority of EC users vaped occasionally: 712 of our 5353 participants were using ECs at at least one of the two time points, including 89 users at both. Initiation, relapse and cessation of smoking Across almost all CC-smoking status, those who had not used ECs at t1 and t2 had the lowest adjusted odds ratios (AORs, Table 2). One exception was the never-smokers at t1 who had already vaped at t1, but they were only few (n = 11), which was one reason why t1-only vapers were grouped with persistent vapers (at t1 and t2). The two groups (t1-only vapers and persistent vapers) could only be separated among persistent smokers. Signi cantly higher AORs for smoking were found for participants starting EC use at t2. If a subgroup of persistent smokers was selected, namely daily-smokers at t1, then a small, but non-signi cant bene cial effect was found (AOR = 0.72, 95% CI [0. 40, 1.30]). This group exhibited dual-use at t1 and had stopped EC use at t2. Remarks: AOR = adjusted odds ratio, adjusted for age, education and language; * additional adjustment for the Fagerström Test for Nicotine Dependence score; persistent t1 daily smokers: smokers at t0 and t1, with daily smoking at t1.
A sensitivity analysis examined almost-daily EC use at: a) either t1 or t2; and b) t1 independent of the use frequency at t2 (  Smoking reduction among persistent smokers at t1 and t2.

t2 smokers' attempts to quit
Negative binomial models demonstrated that, among all smokers, those who were using ECs at t2, independently of whether they used them at t1, had a signi cantly higher incidence rate ratio of the number of attempts to quit smoking. However, when testing occasional and daily smokers at t2 separately, this effect was only signi cant for daily smokers (Table 5). Subgroup of smokers with a good prognosis Finally, we looked at a group of vapers with a good prognosis for reducing and ceasing smoking. These were smokers at t1 using nicotine liquids in at least 2nd generation ECs and with the motivation to reduce or cease CC use. Vapers at t2 were strati ed into four groups (no use vs EC use at t1, and occasional use vs almost-daily EC use at t2) and compared to never-vapers. Models for smoking cessation did not converge for occasional EC users at t2 because they were all still smokers at t2. For almost-daily vapers at t2 who had been non-vapers at t1 (n = 35), the AOR for smoking at t2 was 1.42 (95% CI [0.56, 3.52]; p = 0.44), but it was 0.63 (95% CI: 0.20, 2.02) for those who had already been vapers at t1 (n = 13).
With regard to the number of cigarettes smoked weekly by persistent smokers, almost-daily EC users at t2 who had been non-vapers at t1 smoked signi cantly fewer weekly cigarettes (-22.92 + 1.62 = -21.30) than never-vapers, but this was not true for occasional vapers at t2. Similar reductions were found for those vaping at t1 and occasionally vaping at t2 (-25.23 + 1.62 = -23.62), but not for those with almost-daily use at t2 (Table 6). There were more attempts to quit among t2 smokers for all ever-vapers compared to never-vapers, which were nonsigni cant for persistent vapers with almost daily vaping at t2. Good prognosis: 2nd generation-type ECs, use of nicotine liquids, and motivation to reduce or quit CC use.
Estimate: weekly change in number of cigarettes smoked in mixed models and IRR for binomial count model.

Discussion
The present study of the effects of vaping on reducing and ceasing cigarette smoking used a 4-year longitudinal perspective to investigate a general population sample of young men (mean age 21.3 years old at t1, and 25.4 at t2). The results indicated that the more promising ndings from RCTs (13,14) and the inconclusive ndings from observational studies (15)(16)(17)(18) were not necessarily a contradiction. EC users in a general population may be more heterogeneous than the users motivated to quit in RCTs. In our general population as a whole, no signi cant bene cial effects of EC use were found with regards to reducing or ceasing smoking, but there was an increased number of attempts to quit, con rming other studies (52,53). As in those other general population studies, some EC users, particularly regular EC users motivated to reduce or quit smoking, may indeed have bene tted, but they were relatively few compared to all vapers. Thus, although the effects of EC use were sometimes in a bene cial direction (fewer cigarettes smoked weekly or cessation), they were rarely signi cant. As the number of users in some subgroups was rather small, in line with Greenland et al. (54) or Rothman (55), we did not only interpret signi cant effects or claim that non-signi cant effects were not effects at all, but we also looked at the general picture of effects, e.g. whether they were consistently negative or mixed across different user groups.
Participants already vaping at t1 had similar or higher smoking rates at t2 than never-vapers. Vaping may have had a small bene cial effect for daily smokers at t1 who had been smokers at t0 (persistent smokers at t1), but this effect was non-signi cant, concerned 98 of the total 712 vapers and consisted of an unadjusted difference in cessation of 12 3,30). Approximating the suggestion made by Villanti et al. (20), the group with a good prognosis (i.e. daily vaping, long-term vaping, using nicotine liquids in at least a 2nd generation EC and being motivated to reduce smoking or quit) for meeting bene cial outcomes showed similarly bene cial effects towards smoking cessation, but the sample size was even smaller (n = 13).
Never-smokers and ex-smokers at t1 were not more likely than never-vapers to start smoking or relapse if they had already vaped at t1. Thus, once non-smoking has been achieved or maintained while already vaping, taking up smoking or relapsing is unlikely. Again, the number of such vapers was small compared with all vapers (11 never-smokers and 15 ex-smokers at t1). However, those who started vaping later were signi cantly more likely to be smokers at follow-up, i.e. to take up smoking or to relapse. This may con rm other studies on the greater likelihood of relapse (47) or taking up smoking among younger populations (12). However, as a caveat, for never-smoking, non-EC-users at t1 who smoked and vaped at t2, we do not know which habit began rst. Thus, this nding about relatively young people may also be explained by their generally greater risk-taking and shared vulnerabilities (3,56).
Among daily smokers at t1 who were persistent smokers at t2, there was a small, non-signi cant reduction of 5-7 fewer CCs smoked weekly by EC users than by never-EC-users, but only if they were already vaping at t1 and either continued to vape at t2 or ceased vaping at t2. No relative reduction was found among daily smokers at t1 who initiated vaping by t2. This may mean that EC users have to vape for longer before a reduction in CC smoking appears. These results support ndings from a general population study in France (47), but also from RCTs (21), that the reduction of CC smoking among dual users may be rather small. On the one hand, the few occasional smokers at t1 who persistently used ECs at t1 and t2 (n = 17) had signi cantly increased their smoking by 26 CC more per week than never-vapers. This could indicate that EC use may reinforce smoking-like behaviour (6,57), which could result in heavier CC use among occasional smokers. On the other hand, following the recommendations of Villanti et al. (20), some but not all vapers with the best prognosis for our desired outcomes signi cantly reduced CC use by more than 20 cigarettes a week. Even for smokers with a good prognosis, ndings were mixed. Those already using ECs at t1, but only irregularly at t2, reduced their CC use, which was in line with studies showing that long-term EC use may result in a reduction of CC use (58), but probably only if ECs were also reduced. No reduction in use was found for persistent EC users with almost-daily use at t2. Those who did not vape at t1 but did so daily at t2 showed a reduction of more than 20 cigarettes per week. This may suggest that smoking reductions among dual users may be achieved in the short term, but not in the long term (23). Overall, it seems that dual users may get little direct bene t from their EC use as regards their CC consumption (6,41).
Consistent with other studies is the nding that that dual smokers and EC users at t2, regardless of whether they had used ECs at t1, demonstrated a higher number of attempts to quit. The number of attempts to quit was even higher in the subgroup with a better prognosis than the overall sample. The nding concerning more attempts to quit is well documented in the literature, and it is suggested that this may lead to smoking cessation in the future (3). However, increased numbers of attempts to quit without necessarily increased cessation rates have also often been found (39,40), and our ndings supported this.
The present study's ndings come with some caveats. The sample consisted solely of relatively young men-we can say nothing about women. Young people often only experiment with ECs (29) or try it out of curiosity (13), which the present study supported, showing that few vaped daily or almost daily. Young people are commonly at the beginning of their lives as smokers, and there may be much greater bene ts of EC use among older, long-time, heavy smokers with stronger urges, needs and perhaps motivations to quit smoking. With the past twelve months assessment at t1 and t2, and a long, average follow-up period of four years, we could not investigate relapses and uptakes of smoking and vaping in between. However, this has also been true for most studies with shorter follow-ups. When co-occurrences of smoking and vaping begin at the same time during the 12 months before assessment, we cannot investigate which came rst. However, we do know that before t1, EC use was very rare in Switzerland, and we know whether people were smokers before t1. Thus, almost all smokers at t1 had started smoking before their EC use, and we were able to monitor for vapers at t1 to t2-and similarly for smokers and non-smokers with regards to simultaneous or subsequent EC use. We were also able to say something about dual users at t1 and how they compared when they did or did not continue with EC use at t2. The present study overcame some of the problems inherent in other longitudinal studies: we were able to adjust for nicotine dependence and we followed-up with people who had quit smoking, not only treatment failures (13).

Conclusion
Even critics of vaping do not deny that it may have bene ts for some smokers, particularly if newer vaping devices are used, they are used for longer durations and with greater intensity (daily use), and they are used to aid the motivation to quit. The present study supports such ndings. However, for an evaluation of vaping's overall impact, studies at the general population level are needed (21,38), particularly when it comes to policy decisions, such as giving a price advantage to ECs through the higher taxation of CCs (2,11,12,37). Many vapers have no plans to quit but rather use ECs to circumvent smoke-free policies or to maintain a similar habit as part of their daily routine. Thus, they do not t the strict inclusion criteria of RCTs as regards frequency of EC use, type of CC, dose or duration (15,25,57). Thus, the use of RCTs or the selection of subgroups with good prognostic outcomes does not re ect real-world use (26). The present study showed that for young men, although a few may have bene tted from ECs, at the general population level, vaping probably had no bene t and probably even a negative impact on their smoking behaviour.
Abbreviations AOR: adjusted odds ratio; CC: conventional cigarettes; C-SURF: cohort study on substance use risk factors; EC: e-cigarettes; FTND: Fagerström Test for Nicotine Dependence; IRR: incidence rate ratio; OR: Odds ratio; RCT: randomized control trial; SE: standard error Declarations