The efficacy and acceptability of pharmacological monotherapies and e-cigarette on smoking cessation: a systemic review and network meta-analysis

Background and aims Several pharmacological interventions, such as nicotine replacement therapy (NRT), varenicline, and bupropion, have been approved for clinical use of smoking cessation. E-cigarettes (EC) are increasingly explored by many RCTs for their potentiality in smoking cessation. In addition, some RCTs are attempting to explore new drugs for smoking cessation, such as cytisine. This network meta-analysis (NMA) aims to investigate how these drugs and e-cigarettes compare regarding their efficacy and acceptability. Materials and methods This systematic review and NMA searched all clinical studies on smoking cessation using pharmacological monotherapies or e-cigarettes published from January 2011 to May 2022 using MEDLINE, COCHRANE Library, and PsychINFO databases. NRTs were divided into transdermal (TDN) and oronasal nicotine (ONN) by administrative routes, thus 7 network nodes were set up for direct and indirect comparison. Two different indicators measured the efficacy: prevalent and continuous smoking abstinence. The drop-out rates measured the acceptability. Results The final 40 clinical studies included in this study comprised 77 study cohorts and 25,889 participants. Varenicline is more effective intervention to assist in smoking cessation during 16–32 weeks follow-up, and is very likely to prompt dropout. Cytisine shows more effectiveness in continuous smoking cessation but may also lead to dropout. E-cigarettes and oronasal nicotine are more effective than no treatment in encouraging prevalent abstinence, but least likely to prompt dropout. Finally, transdermal nicotine delivery is more effective than no treatment in continuous abstinence, with neither significant effect on prevalent abstinence nor dropout rate. Conclusion This review suggested and agreed that Varenicline, Cytisine and transdermal nicotine delivery, as smoking cessation intervention, have advantages and disadvantages. However, we had to have reservations about e-cigarettes as a way to quit smoking in adolescents.


Introduction
Unquestionably, tobacco smoking is one of the modifiable factors that heavily contribute to the global health burden.According to a global burden of disease study, there will be an increasing number of 7.69 million deaths and 200 million disability-adjusted life-years attributable to tobacco smoking within this decade if interventions are abscent 1.Multiple behavioral and pharmacologic interventions, both in combination and individually, were proved effective and applied in practice (1).Even though previous randomized controlled trials (RCTs) and meta-analyses showed evidence supporting the effectiveness of behavioral interventions in smoking cessation (1,2), their effectiveness is relatively modest compared with approved pharmacological interventions (1).
Current to the date when this study was performed, there were 7 pharmacological interventions widely approved by most countries: nicotine replacement therapy (NRT, including nicotine mouth spray, inhaler, gum, patch, and lozenge), varenicline, and bupropion.However, inconsistent effectiveness reported by RCTs and metaanalysis of the above pharmacological therapies is not excellent enough, and the relapse rate remains high (3).Besides, the relatively high cost of NRT and varenicline also prevent patients who are in low-income classes from approaching such smoking cessation aids (4).It is still important to innovate novel pharmacological interventions for more cost-effective and acceptable aids in assisting smoking cessation.
As a new product with the potential in assisting smoking cessation, the e-cigarette has already shown evidence of effectiveness and non-inferiority to NRTs in assisting smoking cessation from the previous meta-analysis of both RCTs and observational studies (5,6).Cytisine, due to its similar mechanism with varenicline as a selective partial agonist of nicotinic acetylcholine receptors and low cost of production, has also been previously investigated and proved to be effective and globally affordable in assisting smoking cessation (4,7).
This network meta-analysis aims to systemically and quantitatively evaluate and compare the overall effectiveness and acceptance of all above-mentioned interventions.

Search strategies and literature resources
We searched MEDLINE, COCHRANE Library & PsychINFO for RCTs reporting pharmacological monotherapies and/or e-cigarettes (and equivalents) on smoking cessation.Due to the purpose of comparability and consistency of study cohorts, the time of publication was restricted to be from 2011 Jan 1st [in which the first RCT reporting e-cigarette was released (5)] to 2022 May 31st (in which this network meta-analysis was firstly proposed) during searching.Additionally, references to already-published reviews and meta-analyses with a similar topic were also screened for consideration of inclusion.

Eligibility criteria and study selection
Eligibility criteria were proposed before we perform this network meta-analysis.Inclusion of studies was considered if the study met the following: (1) RCTs; (2) reported in English; (3) study cohorts were recruited in a community-based setting; (4) study cohorts had a persistent smoking history; (5) pharmacological monotherapies or e-cigarette (or its equivalent) were used as an intervention in ≥1 study cohort.Furtherly, studies were excluded if: (1) duplicate records; (2) the study cohort was with a major health condition (e.g., cancer, chronic respiratory diseases, heart and vascular diseases, and schizophrenia or bipolar disorder); (3) follow-up of study endpoint was less than 4 weeks; (4) study outcomes (smoking abstinence, prevalent and/or continuous) were not supported by objective evidence (e.g., saliva cotinine, exhaled CO, serum cotinine, urine cotinine).
Two reviewers (L.Qu and S. Xiang) independently searched and selected studies according to the above strategies and criteria, with disagreement resolved by discussion.All citations retrieved from the database were firstly screened for eligibility at Title/Abstract level, and identified studies were furtherly acquired and examined in full text.Forty studies were eligible and included in the final analysis .

Data identification and extraction
We identified three study outcomes for this meta-analysis due to our study interest, which are defined and listed as the following: (1) Prevalent smoking abstinence (PSA): the percentages of the population who currently quit or reduced cigarette use during the follow-up investigation in between 16 and 32 weeks; (2) Continuous smoking abstinence (CSA): the percentages of the population who consistently maintain smoke quitting or reduction from the first to the last follow up the investigation; (3) Treatment drop-out rates (TDR): the percentages of the population who dropped out from the study or lost to follow-up during the treatment period.
Data identification and extraction were performed by 2 reviewers (L.Qu and S. Xiang) independently.Additional to direct data and indirect data used for the calculation of study outcomes, the baseline characteristics of each study were evaluated and extracted: sampling population, age, location, sex, recruitment setting, smoking history, comparisons, pharmacological dosage, duration of exposure, length of follow up, and lab methods measuring smoking abstinence (see as Table 1).

Data analysis
All outcomes were dichotomous variables measured as n/N (%).The odds ratio (OR) of each outcome was pooled for network meta-analysis (NMA).The NMA used the Bayesian method for multiple-treatment to pool the OR, under the assumption: the heterogeneity is independent of the comparative arms being used (48).We calculated the Bayesian 95% confidence interval (which is known as credible interval) to estimate the   (Continued)  range of the OR results (49).As proposed by the previous study (50), the statistical models for NMA were chosen based on a model comparison criterion called the Deviance Information Criterion (DIC), which is the sum of the posterior expectation of the overall residual deviance and the posterior mean of the parameter of interest (50,51).We firstly applied both random and fixed models for each outcome and calculated the DIC of both models, then the model with a lower DIC is chosen if the difference of DIC in each model is considerable (>5), otherwise the fixed model is chosen if the between-study difference of DIC in each model is insignificant (<5).The absolute value of the between-study variance in the random effect model was assessed by Tau 2 ; the heterogeneity of variation across studies was estimated through I 2 statistics.Additionally, we used the node-splitting method to evaluate the local consistency by separating direct evidence from indirect evidence (50,52).To rank the interventions for each outcome, we estimated the posterior distribution of the ranking probability and their corresponding estimated surface under the cumulative ranking curve (SUCRA) (50,52).The SUCRA is an estimated index to show the cumulative rank probabilities for each intervention and simplifies the entire information about treatment ranking into a single number.
As guided by the Cochrane Handbook for Systematic Reviews of Interventions, the internal validity and quality of this systemic review and NMA were evaluated through the aspects of randomization, blinding of intervention allocation and outcome assessment, and incomplete outcome data (51).All data synthesis and statistical analysis were performed in R with the gemtc package: https://github.com/gertvv/gemtc3 Results

Prevalent smoking abstinence (PSA)
Figures 2A, 3A described the network used for the main analyses of PSA, comprising 60 study cohorts and 13,818 participants.
We analyzed the pooled network effect of PSA for all interventions compared with the control group, using both random and fixed effect models initially.The random-effect model was selected for the final report due to significantly lower DIC (112.84 in the random model, 140.56 in the fixed model), indicating a better efficient result.As presented in Table 2, the confidential presentation of results in terms of mean OR with 95% credible intervals (Crl) compared with the control group was summarized.The pooled effects of all nicotinecontaining products (ONN, TDN, EC) as well as buspirone did not exhibit significant superiority over the control group in terms of prevalent smoking abstinence.The CYT and VAR, both demonstrating significant superiority, exhibited similar odds of prevalent smoking abstinence, approximately twice that of the control group.Despite CYT and VAR showing significant superiority over the control group compared to other intervention groups, these two interventions mostly did not exhibit a significantly different odds ratio for PSA relative to other active intervention groups.The only notable significance observed among active intervention groups was in VAR, with approximately 50% higher odds compared to TDN (Table 2 and Figure 2 are shown here).

Continuous smoking abstinence (CSA)
Figures 2B, 3B described the network of CSA, comprising 42 study cohorts and 18,609 participants.Using the same algorithm described before, the random effect model was selected for the final report due to significantly lower DIC (77.67 in random, 140.62 in fixed model).As presented in Table 2, all comparative interventions except for BUP were associated with significant efficacy for the outcome of CSA compared with CTL.Similar to the absolute values of OR in analyses of PSA, VAR (OR 3.02, 95% Crl 1.9-4.81)and TDN (OR 1.83, 95% Crl 1.09-3.17)demonstrated the highest and lowest OR, respectively.Tau 2 in the analyses of CSA was estimated to be 0.24, indicating a moderate variance; and the I 2 was estimated to be 0.00%, indicating that heterogeneity was minimally considerable.As presented in Figure 3B, inconsistency between direct and indirect evidence was observed in the comparison of VAR/CTL, CYT/CTL, and VAR/CYT.Among those inconsistent results, the direct evidence of VAR/CTL (OR 4.23, 95% Crl 2.57-6.77)yielded a positive CSA reduction on Varenicline use, but the direct evidence of CYT/CTL (OR 1.13, 95% Crl 0.508-2.52)and CYT/VAR (OR 1.13, 95% Crl 0.618-2.15)were ambiguous compared with their combined evidence.Comparative loops with e-cigarette (EC/CTL, EC/TDN, and EC/ONN) were exclusively consistent between direct and indirect evidence, and neither superiority nor inferiority was significant in EC/TDN (OR 1.25, 95% Crl 0.59-2.61)and EC/ONN (OR 0.96, 95% Crl 0.45-1.99)comparisons.

Treatment drop-out rates (TDR)
Figures 2C, 3C described the network of CSA, comprising 42 study cohorts and 18,609 participants.Following similar principles as before, we opted for a random-effects model as significantly lower DIC (80.63 in random, 140.62 in fixed model).As presented in Table 2, With the exception of BUP, all intervention groups exhibited significant superiority over the control group in terms of continuous abstinence rates.Among these, VAR, CYT, EC, and ONN showed odds approximately 2-3 times higher than the control group.Noneligible Tau 2 in the analyses of CSA was estimated to be 0.24, indicating a moderate variance; and the I 2 was estimated to be 0.00%, indicating that heterogeneity was minimally considerable.For the outcome measure of CSA, comparisons among active intervention groups mirrored those of PSA, with only VAR demonstrating significant superiority over TDN.

Treatment ranking
As presented in Table 3, we estimated the posterior distribution of the ranking probability and their corresponding SUCRA for all outcomes.Briefly, CYT is quite likely to encourage both prevalent and continuous smoking abstinence but may lead to dropout.VAR is quite likely to encourage prevalent abstinence, is not particularly effective with continuous abstinence, and is very likely to prompt dropout.In contrast, ONN and EC are least likely to prompt dropout and both are more effective than no treatment in encouraging prevalent abstinence.However, TDN is more effective than no treatment in continuous abstinence, with neither significant effect on prevalent abstinence nor dropout rate (More details seen in Table 3)."

Discussion
As far as we know, our study is the first to report the efficacy and acceptability of five major pharmacological monotherapies and e-cigarette on smoking cessation through network meta-analysis including RCT studies.And this NMA including 40 studies found that (1) Varenicline is more effective intervention to assist in smoking cessation during mid-to long-term (16-32 weeks) follow-up, but is not particularly effective with continuous abstinence, and is very likely E-cigarettes and oronasal nicotine are least likely to prompt dropout and both are more effective than no treatment in encouraging prevalent abstinence.Finally, transdermal nicotine delivery is more effective than no treatment in continuous abstinence, with neither significant effect on prevalent abstinence nor dropout rate.
Our findings are consistent with the approach recommended by current mainstream clinical smoking cessation guidelines, such as the use of Varenicline as a first-line pharmacological intervention to assist in smoking cessation by the 2020 American Thoracic Society guidelines (53) and the recommendation of NRT, varenicline and bupropion as first-line pharmacological interventions for smoking cessation in the 2018 ACC Expert Consensus (54).Since e-cigarettes (or equivalent products) have a pharmacological mechanism for distributing nicotine to the body, their potential cessation effect has also gained the attention of manufacturers.This study also showed their similar effects to NRT treatment in terms of smoking cessation effectiveness.
However, our findings should be cautiously interpreted.The ethnic distribution of overall participants involved in this NMA is considerably uneven since most of the included RCTs were performed in Europe and North America.Thus, the results of this NMA may not be generalized to other ethnical groups due to the differences in tobacco dependence and cessation in acculturation and nicotine metabolism levels described in previous studies (55, 56).Gender differences in pharmacotherapies of smoking cessation are also non-negligible since it is clear that certain medication shows different efficacy between male and female participants desiring smoking cessation (56).The selection criteria of smoking intensity and duration in each included study can vary considerably, ranging from light-intermittent to heavy-daily smoking, and such differences in smoking intensity may indirectly affect patients' confidence in Frontiers in Public Health 13 frontiersin.orgquitting smoking (57).A more specifically stratified discussion in participants with different smoking intensities should be investigated in further studies.
Overall, there is a moderate level of variance among all included studies.Such variance may result from several possible aspects.Firstly, interventions were artificially classified, and the oronasal nicotine replacement therapy includes four different FDA-approved pharmacotherapies (nicotine nasal spray, nicotine inhaler, nicotine gum, and nicotine lozenge) with possibly variant effectiveness due to different nicotine delivering dosages and delivering routes.Secondly, the overcall control group also has the potential of being part of the variance.We summarized all placebo groups and control groups without blind settings from each trial as one single group, and there may also be differences in the effect on treatment outcomes between the different placebo and the unblinded control settings.Additionally, it has been described by Chan 2021, etc. that the diversity of e-cigarette products may also be problematic to generalize the results to newly-created e-cigarette products (5).And last, of all, the method chosen for verification of outcome measurement is also concerning.Though serum and urine cotinine are used for smoking cessation verification and quantitative measurement in some trials, exhaled CO is the mostly applied biochemical method for the same purpose and has revealed several shortages, including short half-life (58), and false-positive results with other smoking products (e.g., cannabis) (59).
It has been proved by several previous meta-analysis and RCTs and has been validated by this NMA that e-cigarette is effective in assisting smoking cessation.Compared to PSA, E-cig had a higher probability of superior ranking in smoking cessation effectiveness as measured by CSA, and this similar finding is also observed in the oronasal nicotine group.Based on this finding, we hypothesized that patients who use e-cigarettes and oronasal nicotine products would However, we should still be cautious to approve e-cigarettes as a therapeutical intervention for smoking cessation.E-cigarette or vaping product use-associated lung injury (EVALI), a novel entity including a broad spectrum of pulmonary diseases and may lead to respiratory failure, has continuously been reported (60-66).There is also growing evidence indicating generalized pulmonary toxicity may be caused by inhaling electronic cigarette vapor (67).Additionally, the psychoactive substances and special flavors of vapor have led to a surge in usage, especially among adolescents (68).A study from the U.S. indicates that more than 40% of high school students have tried e-cigarettes in the past year in 2020 (69).What is more alarming is that studies have proved that initial e-cigarette use is also associated with subsequent cigarette smoking initiation among adolescents and young adults (70,71).Further discussion on whether e-cigarettes can be used as a pros-outweigh-cons intervention in assisting smoking cessation should follow more investigations on their long-term safety.On balance, we have reservations about e-cigarettes as a way to quit smoking.Inconsistency check between direct and indirect evidence in the network meta-analysis of PSA (A), CSA (B), and TDR (C).PSA, prevalent smoking abstinence; CSA, continuous smoking abstinence; DOR, drop-out rate; TDN, transdermal nicotine; ONN, oronasal nicotine; VAR, varenicline; BUP, bupropion; CYT, cytisine; CTL, controls; 95% Crl, Credible interval.

Conclusion
Our study reported the efficacy and acceptability of five major pharmacological monotherapies and e-cigarette on smoking cessation through network meta-analysis including 40 RCT studies.We recommended that Varenicline, Cytisine and transdermal nicotine delivery, as smoking cessation intervention, have advantages and disadvantages.However, we had to have

FIGURE 1 Flowchart
FIGURE 1Flowchart of study inclusion.

2
FIGURE 2Network meta-analysis of eligible comparisons.(A) Network meta-analysis for comparisons of prevalent smoking abstinence (blue).(B) Network metaanalysis for comparisons of continuous smoking abstinence (green) and (C) Network meta-analysis for comparisons of treatment drop-out rates (orange).n indicates the number of total participants.

TABLE 1
Baseline characteristics of each identified studies.

TABLE 1 (
Continued) there is no clear evidence that either varenicline or flexible, dual-form NRT increase quit rates in the long-term when compared to NRT monotherapy.(Continued)

TABLE 3
Posterior distribution of the ranking probability and the surface under the cumulative rank curve (SUCRA) for each treatment in network meta-analysis.