Key points
Validated smoking cessation pharmacotherapies and e-cigarettes can induce sleep disturbances with different degrees of frequency.
Our network meta-analysis shows a more favorable profile of nicotine substitutes for insomnia and bupropion for parasomnia.
It seems essential to systematize the assessment of sleep disturbances in the initiation of smoking cessation treatment. This could help professionals to personalize the choice of treatment according to sleep parameters of each patient.
Introduction
Tobacco has a very detrimental impact on public health, killing up to 50% of its users.1 Its consumption causes a complex dependence and has multiple harmful consequences, with various neoplastic, cardiovascular, and respiratory diseases, inflicting a high cost on society.1–Reference Degenhardt, Charlson and Ferrari3 Despite declining smoking prevalence in many countries, there are disparities among vulnerable patients, young people, and women.Reference Perriot, Underner and Peiffer4–Reference Piper, Cook and Schlam7
Current treatment ranges from minimal counselling to pharmacological treatments and cognitive behavioral therapies (CBT). Validated pharmacotherapies include nicotine replacement therapy (NRT), varenicline, and bupropion.Reference Cahill, Stevens and Lancaster8 They increase the chances of smoking cessation,Reference Hartmann‐Boyce, Chepkin, Ye, Bullen and Lancaster9, Reference Cahill, Lindson‐Hawley, Thomas, Fanshawe and Lancaster10 but many studies show a high relapse rate in the long term.Reference Marlatt and Donovan11, Reference Yeomans, Payne and Marton12 Electronic cigarettes (e-cigarettes) are also part of the emerging smoking cessation methods since the 2010s, with frequent use among smokers.Reference Farsalinos, Poulas, Voudris and Le Houezec13, Reference Levy, Borland and Lindblom14 A recent meta-analysisReference Hartmann-Boyce, McRobbie and Butler15 reported a significant efficacy of the electronic cigarette with nicotine versus placebo in terms of cessation and reduction after 6–12 months, but its safety is highly debated due to insufficient good-quality randomized controlled trials.
There is an important variability in treatment response, and one current challenge is to identify the causes of treatment failure to move toward personalized management. For example, evidence suggests that sleep disorders can be important for smoking cessation.Reference Benowitz16
First, cigarette smoking can alter sleep architecture, and current smokers experience greater difficulty initiating and maintaining sleep.Reference Gibson, Munafò, Taylor and Treur17–Reference Peters, Fucito, Novosad, Toll and O’Malley21 Acute nicotine intake from cigarette smoking stimulates the release of key neurotransmitters that regulate sleep architecture. In animal studies, nicotine stimulates serotonin release in the dorsal raphe nucleus, which contribute to suppressing the ponto-geniculo-occipital spike of the last stage of sleep and the rapid eye movement (REM) sleep, which is important for memory and spatial consolidation.Reference Guzmán-Marín, Alam, Mihailescu, Szymusiak, McGinty and Drucker-Colín22–Reference Saint-Mleux24 Other studies have shown a dose-dependent effect of nicotine on REM sleep: a lower dose stimulates REM sleep, while a higher one suppresses it and reduces sleep time.Reference Jaehne, Loessl, Bárkai, Riemann and Hornyak25 Saint Mleux et al. found that nicotine inhibits key regions implicated in promoting sleep via activation of norepinephrine release.Reference Saint-Mleux24 In humans, the Zhang studyReference Zhang, Samet, Caffo and Punjabi18 shows that smokers have a longer stage 1 sleep phase and a higher percentage of stage 2 sleep (light sleep), decreasing sleep quality. Other studies reported that smokers are more vulnerable to longer sleep latency, more awakening, and a shorter sleep time.Reference Jaehne, Loessl, Bárkai, Riemann and Hornyak25, Reference Patterson, Grandner, Malone, Rizzo, Davey and Edwards26
Second, sleep disorders are an important part of withdrawal symptoms.Reference Boakye, Wyse and Morales-Celis19–Reference Peters, Fucito, Novosad, Toll and O’Malley21, Reference Ashare, Lerman and Tyndale27 For example, 42% of smokers report insomnia during abstinence,Reference Cummings, Giovino, Jaén and Emrich28 and sleep disturbances increase following smoking cessation. Most of these disorders disappear after three months. For smoking cessation outcomes, smokers with prior sleep disorders have shown a lower success in later smoking cessation attempts. Moreover, sleep quality at the beginning of the cessation attempt predicted relapse.Reference Patterson, Grandner, Malone, Rizzo, Davey and Edwards26
Third, sleep disorders are important side effects of validated pharmacotherapies. In a meta-analysis, up to 10% of participants treated with NRT reportedly experience insomnia that can persist more than 12 weeks after stopping.Reference Jaehne, Loessl, Bárkai, Riemann and Hornyak25 At the beginning of smoking cessation treatments, up to 50% of smokers report sleep disturbance. According to Paterson et al.Reference Patterson, Grandner, Malone, Rizzo, Davey and Edwards26, 4–21% of sleep disorders with bupropion and up to 46% of varenicline-seeking smokers reported difficulty sleeping and abnormal dreams.
To our knowledge, no recent meta-analysis or systematic review on sleep disturbances and smoking cessation treatment exists. Due to the limitations of existing systematic reviews and the emergence of new cessation methods, an update on this topic seems necessary, and more precisely on the occurrence of insomnia and parasomnia. Indeed, according to the International Classification of Sleep Disorders (ISCD) 3,Reference Sateia29 sleep disorders commonly reported in smoking cessation studies can be classified as insomnia (difficulty to initiating and maintaining sleep) and parasomnia (abnormal dreams, nightmare). Using a systematic review and network meta-analysis, we aim to evaluate the insomnia and parasomnia induced by validated smoking cessation pharmacotherapies and e-cigarettes.
Network meta-analysis allows a comparison of several health interventions for a given indication. It combines direct evidence (treatments compared two by two) with indirect evidence (treatments compared via a common comparator).Reference Salanti, Higgins, Ades and Ioannidis30, Reference Riley, Jackson and Salanti31 This analysis allows for a more accurate estimate and can establish a relative ranking between treatments for the desired endpoint.Reference Salanti, Ades and Ioannidis32
Materials and methods
We conducted a systematic review following the Preferred Reporting Items for Systematic review and Meta-Analyses (PRISMA) guidelines.Reference Moher, Liberati, Tetzlaff and Altman33
Research strategy
The databases Cochrane Central Register of Controlled Trials (CENTRAL), Clinicaltrials.gov, and PubMed were consulted until 11 of august 2022 after a first exploratory research. We used the International Classification of Sleep Disorders (ISCD 3) and MedRa classification to define our research strategy.Reference Sateia29, 34
The keywords used are divided into two parts:
-
1. #1 For smoking cessation methods:
-
a. MeSH: Tobacco use cessation devices, Electronic nicotine delivery systems, Electronic nicotine delivery device, Varenicline, Bupropion, Electronic cigarettes, Smoking cessation agents
-
b. Non-MeSH: Nicotine replacement therapy, e-cigarettes
-
-
2. #2 For sleep disorders:
-
a. MeSH: Sleep–wake disorders, Sleep apnea syndromes, Parasomnia, Restless legs syndrome, Sleep initiation and maintenance disorders, Dyssomnias, Insomnia, Disorders of initiating and maintaining sleep
-
b. Non-MeSH: Sleep disorders, abnormal dreams, sleep disturbance
-
A manual search was conducted: We selected Cochrane reviews which reported sleep outcomes from 4 meta-analysis on NRT (2016 update), varenicline (2018 update), bupropion (2020 update), and electronic cigarettes (2021 update).
Then, we used two research equations:
-
- A PubMed broader research from 2016 (date of the latest Cochrane update) to 08/11/2022: #1 AND Smoking Cessation, Filter: Randomised controlled trial
-
- A narrow specific research: #1 AND #2
Data selection
We included English-language literature randomized controlled trials double-blind, single-blind, or open-label.
The eligibility criteria following the PICO model (Patient/Population, Intervention, Comparison, Outcomes) were:
-
- Patient/Population: adult smokers (men and women over 18) without unstable comorbidity and without pregnancy. Patients with psychiatric or addiction comorbidities were excluded.
-
- Intervention: validated therapeutic smoking cessation methods and the electronic cigarette with a duration of at least 1 month.
-
- Comparison: active, placebo, or no treatment interventions.
-
- Outcomes: The primary outcome is determined by the presence of insomnia and/or parasomnia (including abnormal dreams and vivid dreams).
Screening and data extraction
Two authors (CP, PV) independently screened the titles and abstracts of search hits to select studies of interest and reviewed the full text with the Covidence software.35 Disagreements were resolved by discussions between the authors and with a third view (CL). Information on methodology, participants, interventions, and outcome measures was collected by CP on Excel spreadsheets and cross-checked by PV and CL. If the outcomes measures were not prespecified in the studies, the authors searched on the side effects reported in the full texts.
Evaluation of the quality of studies
The risk of bias was assessed using the Cochrane collaboration risk of bias (RoB 2) tool.Reference Sterne, Savović and Page36 We assessed the randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, selection of reported results (with respect to prespecified analysis), and overall bias. Judgments of risk were classified as low, high, or of some concern.
Quantitative analysis
We used the “Meta insight V 3.19” website using the WinBUGS tool and Revman 5.4 software.Reference Owen, Bradbury, Xin, Cooper and Sutton37, 38 Analyses were stratified for each outcome criterion and by intervention.
We used a Mantel–Haenszel (MH) method for pairwise analysis with a random-effect model. The results are presented as relative risk (RR) for binary variables with a 95% confidence interval.
To obtain a significant result, we chose a p-value <0.05 and a confidence interval not including 1. A relative risk greater than 1 indicates a negative effect on sleep. Heterogeneity was assessed using the I2 statistic. An I2 estimate of >50% corresponds to substantial heterogeneity, moderate heterogeneity to 25–50%, and low heterogeneity when it is <25%.
For the network meta-analysis, we used a Bayesian method. The different interventions and placebo mapping were represented by a network plot for the two analyses.
Transitivity was maintained by selecting studies with similar indications for the interventions, for example, smoking cessation.
The consistency of the network, corresponding to the absence of disagreement between the results of the direct and indirect comparisons, was assessed by a global inconsistency test.
The analysis was not preregistered, and the results should be considered exploratory.
Results
Selection of studies
We identified 1261 articles using our search strategy. After removing duplicates and screening titles and abstracts, 328 full texts were assessed for eligibility. Two hundred ninety-one studies were excluded mainly for lack of outcome data or inappropriate study design, setting, and wrong outcomes. Finally, 37 studies were selected for the quantitative analysis (Figure 1).Reference Ahluwalia, Harris, Catley, Okuyemi and Mayo39–Reference Zhang, Mansoursadeghi-Gilan and Hussain75
Characteristics and quality of the studies
Table 1 shows the characteristics of each study. They date from 1993 to 2022. Participants were smokers with no comorbidity or psychiatric history.
In blue: Interventions not included in the analysis.
Abbreviations: CBT, cognitive behavioral therapies; NRT, nicotine replacement therapy; 7 DPP, 7 day point prevalence abstinence.
The interventions found were varenicline (23 studies), bupropion (15 studies), and nicotine replacement therapy (10 studies). No studies on electronic cigarettes could be included in the quantitative analysis. Comparators were placebo (28 studies), active treatment (nine studies) or behavioral therapy (one study). Eight trials used three or more arms (Table 1). Intervention duration ranged from 4 weeks (prior to quit date) to 18 weeks post-quit date. Follow-up times ranged from 12 to 52 weeks. Most of the studies have the same endpoint of smoking abstinence with different parameters (7-day point prevalence or continuous abstinence, confirmed by exhaled CO, saliva cotinine, or urinary anabasine concentration), and two studies use a primary outcome focus on adverse effect.Reference Anthenelli, Benowitz and West40, Reference Williams, Anthenelli and Morris74 The full text analysis identified insomnia outcomes in 34 studies and parasomnia in 23. Twenty-two studies reported both of the outcomes.
Of the 37 studies, 18 were classified as low risk of bias, 13 as some concerns risk, and six as high risk (Figure 2).Reference Riley, Jackson and Salanti31
Data analysis
Pairwise meta-analysis (direct comparisons)
For the analysis focused on the outcome “Insomnia,” 34 studies were included.
Subgroup effects analysis (Figure 3) showed that insomnia was significantly more frequent with varenicline (RR: 1.54 [1.30–1.81]) and bupropion (RR: 1.86 [1.63–2.13]) than with placebo. On the other hand, varenicline is significantly less responsible for insomnia than bupropion (RR: 0.73 [0.64–0.84]). Bupropion caused insomnia significantly more frequently than NRT (RR: 1.41 [1.18–1.68]). Other comparisons are non-significant. There was substantial heterogeneity in two comparisons: varenicline versus (vs) placebo, varenicline vs NRT. Bupropion vs placebo and NRT vs placebo resulted in moderate heterogeneity. Varenicline vs bupropion and bupropion vs NRT are homogeneous (Figure 3).
For the analysis focused on the endpoint “Parasomnia,” 23 studies were included.
The analysis of subgroup effects (Figure 3) shows that varenicline and NRT caused significantly more parasomnia than placebo (RR: 2.42 [1.75–3.36] and RR: 3.46 [1.67–7.15], respectively). However, the subgroups effects of bupropion vs placebo and varenicline vs NRT were not significant. Parasomnia were significantly more frequent with varenicline compared to bupropion (RR: 1.55 [1.06–2.26]). Parasomnia were less frequent with bupropion than with NRT comparing them directly (RR: 0.35 [0.21–0.59]). The varenicline vs placebo comparison showed substantial heterogeneity, and only one comparison had low heterogeneity: bupropion vs placebo (Figure 3).
Network meta-analysis (Indirect comparisons)
The network structures with “Insomnia” and “Parasomnia” outcomes are available in the supplementary material file (Supplementary Figure S1).
For insomnia, Table 2 shows the results based on a Bayesian network meta-analysis. The cessation methods significantly increased the risk of insomnia. NRT was significantly less harmful to insomnia than bupropion and varenicline (RR: 0.62 [0.49–0.76] and RR 0.79 [0.64–0.95]). Varenicline had a lower risk of insomnia than bupropion (RR:0.78 [0.66–0.92]) (Table 2). In the ranking probability analysis, placebo and NRT had better profiles (Supplementary Figure S2).
In the lower left triangle, comparisons should be read from left to right. In the upper right triangle, comparisons should be read from right to left (that is treatment 1 versus treatment 2). Significant values are in bold (confidence interval not including 1).
The Bayesian method reveals that bupropion caused significantly less parasomnia than NRT (RR: 0.52 [0.32–0.82]) and varenicline (RR: 0.59 [0.41–0.86]). Comparisons between varenicline and NRT were not significant (Table 3). The rank probability analysis shows a better profile of bupropion than varenicline and NRT (Supplementary Figure S3).
In the lower left triangle, comparisons should be read from left to right. In the upper right triangle, Comparisons should be read from right to left (that is treatment 1 versus treatment 2). Significant values are in bold (confidence interval not including 1).
Coherence analysis
A consistency analysis is performed by comparing the values of the direct and indirect comparisons (Supplementary Tables S1 and S2). A p-value greater than 0.05 means that there is no statistically significant difference. Here, all the comparisons for insomnia have a p-value greater than 0.05; the results are therefore consistent.
For parasomnia analysis, two comparisons are inconsistent: NRT vs varenicline, and bupropion vs placebo.
Discussion
This network meta-analysis is based on 37 studies with 25 011 patients randomly assigned to five different interventions or placebo for smoking cessation. The original study confirms that insomnia and parasomnia are more frequent with smoking cessation therapies than with placebo except bupropion. There is a more favorable profile of NRT for insomnia and bupropion for parasomnia than with other smoking cessation treatments. Network meta-analysis is a validated and recognized systematic scientific method. Its value is based on a good level of evidence.Reference Riley, Jackson and Salanti31, Reference Salanti, Ades and Ioannidis32, Reference Li, Threapleton and Wang76, Reference Panikker, Lord and Jarman77 To our knowledge, it is the first network meta-analysis on the topic of insomnia and parasomnia induced by pharmacotherapies for smoking cessation with a ranking of the different smoking cessation methods to guide the choice of treatment.
For NRT, in non-smokers, transdermal nicotine intake reduced REM sleep with a complete recuperation after stopping. In smokers, Gourlay et al.Reference Gourlay, Forbes, Marriner and McNeil78 found that sleep disorders with NRT are correlated to the severity of nicotine dependence and are more frequent in women. Frederickson found a correlation between plasma cotinine levels and the severity of sleep disorders.Reference Fredrickson, Hurt and Lee79 In withdrawal periods, NRT increases arousal and reduces sleep time.Reference Jaehne, Loessl, Bárkai, Riemann and Hornyak25, Reference Wolter, Hauri and Schroeder80 According to Vasquez et al., transdermal nicotine can disrupt, like cigarette smoking, the PGO activity in cats.Reference Vazquez, Guzmán-Marín, Salín-Pascual and Drucker-Colín81 The time administration is important to consider: The 16 h nicotine patch reduces parasomnia but contributes to a night craving related to a fall in nicotine concentration. Compared to 16 h nicotine patch, there is less microarousal and an increase of the REM period with 24-hr nicotine patches.Reference Page, Coleman and Conduit82, Reference Staner, Luthringer, Dupont, Aubin and Lagrue83
In our study, varenicline increases the number of awakenings and reports of parasomnia compared to bupropion. This effect for varenicline is confirmed by polysomnographic studies.Reference West, Baker, Cappelleri and Bushmakin84, Reference Pataka, Frantzidis and Nday85 For bupropion, the effects on sleep architecture are unclear, with few studies available. These two drugs have different actions. Bupropion is an antidepressant and acts by inhibiting the dopamine reuptake in the brain reward center. As a partial agonist of alpha4beta2 nicotinic acetylcholine receptors, varenicline stimulates dopamine release and blocks the action of nicotine cigarette intake. The important rate of sleep disorders frequency with varenicline in our study can be linked to a nicotinic disturbance and dopamine dysregulation, which can be implied in parasomniaReference Happe and Berger86
However, our study has limitations. First, we define the selection criteria, including healthy smokers without comorbidity and those who were not hospitalized. This choice allows us to avoid confounding factors that can affect sleep quality and increase insomnia and parasomnia.Reference Chakravorty, Chaudhary and Brower87–Reference Lewis89 Nevertheless, smoking cessation in patients with comorbidities, particularly psychiatric or co-addictions, remains a public health issue. A network meta-analysis including these different selection criteria would be interesting to carry out.
Second, some cessation methods are more extensively analyzed and have had longer follow-up periods, while others are understudied. For example, the results reported for nicotine substitutes or varenicline are numerous, while no studies on electronic cigarettes could be included in our analysis. Most of them do not report sleep disturbances or do so in an imprecise manner. In addition, studies on smoking cessation are not systematically published, making their inclusion and integration problematic. This difference in data availability could potentially create a selective reporting and publication bias.
Most of the studies analyzed were sponsored by the manufacturers of varenicline and nicotine replacement products. Previous work on nicotine replacement therapies has shown that industry-sponsored trials are significantly more likely to have favorable results than independent trials.Reference Etter, Burri and Stapleton90 However, most of the studies reviewed here are of high-quality evidence (randomized controlled trials) and have a low risk of bias.
There is also heterogeneity in the definitions of sleep disorders. Insomnias are usually explored in withdrawal scales such as the Minnesota Tobacco Withdrawal Scale (MNWS),Reference Hughes and Hatsukami91 and parasomnias are mainly reported as abnormal and vivid dreams in the side effects reported. These outcomes were mainly not prespecified in most of the studies and were extracted in the side effects reported. This constitutes a selection bias. We used the definitions for the ISCD 3;Reference Sateia29 other sleep dimensions could not be extracted.
However, using a single coding scheme for future randomized trials would provide consistency of outcome and limit this measurement bias. For example, these future trials could use a standardized questionnaire such as the Insomnia Severity Index (ISI) or the Pittsburgh Sleep Quality Index (PSQI).Reference Morin92, Reference Buysse, Reynolds, Monk, Berman and Kupfer93 A systematic collection of sleep disorders in future studies would reduce the inconsistencies between direct and indirect estimates. Nevertheless, our study’s transitivity was respected, strengthening our analysis. Indeed, we selected only studies with an identical intervention indication: smoking cessation.
It seems difficult to distinguish sleep disorders related to withdrawal symptoms from the side effects of pharmacological treatment. However, the persistent disturbances observed in patients on NRTReference Mills, Wu, Lockhart, Wilson and Ebbert94 and our results have shown a higher frequency of sleep disturbances with pharmacological treatments compared to placebo. This effect suggests specific mechanisms, but little is known in the literature and could not explain the difference highlighted in this meta-analysis.
If smoking cessation is a factor for improving sleep health, poor sleep quality can reduce the success of cessation.Reference Short, Mathes, Gibby, Oglesby, Zvolensky and Schmidt95 Attention to sleep patterns before starting treatment and considering the side effects on sleep associated with smoking cessation therapy are relevant to increasing smoking cessation probability.Reference Dorner, Tröstl, Womastek and Groman96-Reference Okun, Levine, Patricia, Perkins and Marcus98 Thus, health professionals could promote sleep hygiene measures and adjust dosages to prevent the onset or worsening of sleep disorders. Informing the patient of the links between smoking cessation and sleep health would make it possible to include them in the choice of method and thus make them an actor in the abstinence process.
Conclusion
In conclusion, validated smoking cessation pharmacotherapies can induce sleep disturbances with different degrees of frequency. Our network meta-analysis shows a more favorable profile of nicotine substitutes for insomnia and bupropion for parasomnia. However, our results are qualified by the presence of inconsistencies. These are probably due to a lack of homogeneity in the selected studies and data analysis of specific interventions.
Our study is innovative and deals with a current problem. Current management is increasingly aimed at refractory and anxious smokers who often suffer from sleep disorders. Network meta-analysis—an emerging, validated and recognized method applied to these issues—contributes to scientific research.
Systematizing the assessment of sleep disorders in the initiation of smoking cessation seems essential. This could help health professionals in supervising smoking patients to adapt their practice. Furthermore, considering co-addictions, broadening the populations studied (such as patients with psychiatric comorbidities), and standardizing the measurement are additional avenues for future research on this subject.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1092852924000087.
Author contribution
Conceptualization: C.P., P.D., P.V.; Formal analysis: C.P., P.V.; Writing – original draft: C.P., C.l., N.J., P.D., P.C., P.V.; Writing – review & editing: C.P., C.l., N.J., P.D., P.C., P.V.; Supervision: C.l., N.J.; Validation: C.l., P.C.; Visualization: C.l.; Methodology: P.V.
Financial support
This research did not receive grants from any funding agency in the public, commercial, or not-for-profit sectors.
Disclosure
The authors declare none.