Smoking and Incidence of Insomnia: a systematic review and meta-analysis of cohort studies

Background: Both smoking and insomnia are worldwide problems and this study aims to investigate the impact of smoking on the incidence of insomnia. Methods: PubMed, EMBASE and OVID were searched through March, 2020. Cohort studies reporting the effect of smoking on the incidence of insomnia were included. We quantitatively analyzed the basic framework and study characteristics, and then pooled estimate effects with 95% condence intervals (CIs) of outcomes of each included studies using xed-effects meta-analyses. Results: This systematic review included six cohort studies involving 12445 participants. Quantitatively summarized results suggested smoking could signicantly increase the incidence of insomnia (OR: 1.07, 95%CI: 1.02,1.13). Regular smoking was signicantly associated with incidence of insomnia (OR=1.07, 95% CI:1.01,1.13). As for occasional smokers and ex-smokers, the pooled analysis didn’t indicate a signicant association (occasional smoker: OR=2.09, 95% CI:0.44,9.95; ex-smoker; OR=1.02, 95% CI:0.67,1.54). Subgroup analysis by age, gender ratio and region showed statistically signicant relationship between smoking and incidence of insomnia in specic groups. Conclusions: Integrated longitudinal observational evidence identied smoking as a signicant risk factor of insomnia. Considering the limited amount of available studies, more high-quality and prospective cohort studies of large sample sizes are needed to explore details of this association.

Although previous studies suggested that smoking might be a contributing factor for insomnia, there were considerable differences among the results of studies [14,19,20]. Besides, most of the studies were cross-sectional studies, which couldn't re ect the sequential order and causal relationship between smoking and insomnia. We found several recent reviews which focused on this project but no metaanalysis has been done [21][22][23]. Therefore, our study focused on the association between smoking and incidence of insomnia of cohort studies. Besides, it has been reported that demographic factors including gender, age might be associated with insomnia [24,25] and the degree of correlation between smoking and insomnia varied with region [1,26], which were discussed in our study through subgroup analysis.
Smoking status were also discussed for more speci c results and guidance to public health management.

Methods
This work was performed according to the MOOSE (Meta-analysis Of Observational Studies in Epidemiology) guidelines [27].

Literature search
We identi ed relevant studies published from 1946 to March 2020 by searching on PubMed, EMBASE and OVID. Studies were restricted to human species. Following search keywords were used for searching: (1) smoking, smoker, cigar and their synonyms; (2) insomnia and its synonyms. No lter for study designs was applied. Pertinent studies were also identi ed during the screen of reviews and reference lists of search results. After the exclusion of duplications, titles and abstracts were screened during the initial selection. Further review was based on the full text. NH and CYW screened all the articles independently. Their results reached consensus. A senior reviewer (SYC) would check the article and make decisions if there was disagreement. (Figure 1)

De nition and Study selection
We included studies that t DSM (The Diagnostic and Statistical Manual of Mental Disorders) -5 for diagnostic criterion of insomnia: di culty initiating or maintaining sleep in addition to impaired daytime functioning or clinically signi cant discomfort because of sleep problems [28]. And literature included must provide the diagnostic standard which has been tested or used by other studies or acknowledged by public. The diagnostic standards are provided in Table 1. Besides, according to DSM-III-R, insomnia symptoms were divided into several types which were further discussed in our study: (a) Di culty getting to sleep; (b) Wake up during the night and have a hard time getting back to sleep; (c) Wake up repeatedly and can't get back to sleep; (e) The feeling of not having enough rest during the day, no matter how many hours of sleep you had. For each symptom, responders indicated the frequency with which it was experienced: very frequent (16 or more days or nights per month), frequent (5-15 days or nights per month), infrequent (2-4 days or nights per month), rare (1 or fewer days or nights per month).
Smokers are people who have experience of smoking or exposure to cigarette or tobacco. Different types of smoking status are classi ed by smoking period and number of tobaccos people smoke/smoked each day. A non smoker is de ned as a person who never smokes and ex-smoker is people who ever smoked but stopped when the data was collected. Regular smokers are people who started smoking before and stayed at a level of daily smoking at the endpoint. According to the time of initiation, regular smokers are divided into two types: heavy/continuous smokers are people who started smoking early, reached their maximum level (i.e., about one pack a day or more) in their late twenties, and then maintained that level into their thirties (the endpoint). Late starters are people who started smoking in late adolescence but reached the same level of smoking (i.e., about one pack a day) as the heavy/continuous smokers in the late twenties and then maintained that level till the endpoint. The occasional smokers are people who increased the amount of smoking from adolescence to the early twenties and then stayed at a level of less than daily smoking during adulthood.
Studies were included if they satis ed the following criteria: 1. Cohort studies. 2. Studies designed to examined the relationship between smoke exposure and incidence of insomnia among healthy populations. 3. Comparators de ned as nonsmokers or never smokers. 4. Studies that provide clear hazard ratio(HR) or odds ratio(OR) with 95% con dence interval(CI). The exclusion criterion included: 1.
Duplications or same cohorts with similar study period. 2. Studies which did not provide clear hazard ratio(HR) or odds ratio(OR) with 95% con dence interval(CI), and could not retrieve by contacting with the authors. 3. Studies which restricted the population to speci c patients, such as patients with depression at the baseline.
The study was independently screened by two authors (NH and CYW) using the criteria described above and the different opinions encountered in the study screening would be discussed throughout the process until authors reached a mutual consensus.

Data Extraction
Data extraction was c0onducted by two reviewers (NH and CYW) independently. From each eligible study, we collected authors, publication year, geographical region, gender ratio (women/men), mean age, smoking status, study period, study design, sample size, endpoints, diagnostic criteria insomnia, HRs or ORs with 95% CI. While there were results of different calculation models, we extracted the HR or OR fully adjusted for covariates. If necessary, we tried to contact the corresponding author of relevant articles for information.

Quality assessment
Used Newcastle-Ottawa Quality Assessment Scale (NOS) (http://www.ohri.ca/ programs/clinicalepidemiology/oxford.asp) as reference according to our included study design, we assessed the quality of the eligible studies from 3 individual perspectives: the population selection; the comparability in study design; and the ascertainment of outcome or exposure. Two reviewers (NH and CYW) performed the assessment separately and a senior reviewer (SYC) resolved disagreement. (Table 1 and Supplementary  Table 1)

Statistical analysis
Because of low heterogeneity (I 2 =14.9%, p=0.310), we chose xed-effects model to calculate the pooled HR/OR and 95% CI. Heterogeneity was tested with the I 2 statistic. We combined Funnel plot asymmetry and Egger's regression to detect publication bias, and de ned signi cant publication bias as a p value < 0.1. Sensitivity analysis was used to examine the stability of the pooled results based on the results of quality assessment, whereby each study was omitted at a time and recalculated the pooled OR of the rest studies. We also performed subgroup analysis by age, gender ratio, region and smoking status to examine details about the association. Meta-analysis was performed with Stata software (version 16.0; College station, TX, USA). All p values were two-side with a signi cant level at 0.05.

Literature selection
A total of 592 studies were included in our study, including the results of keywords search (n=586) and additional records identi ed from reference lists of other studies (n=6). After the detailed reading of full texts, ve articles [29][30][31][32][33] reported the relationship between smoking and incidence of insomnia were potentially eligible for meta-analysis ( Figure 1) and one other article [34] was eligible for further discussion as it divided insomnia into different symptoms. The main information of eligible studies was given in Table 1. We assessed the quality of included studies from three aspects and their scores were no less than seven points (total score=seven points) ( Table 1 and Supplementary Table 1).

Bias from character of population
Cohorts of four included studies [29][30][31]33] were representative of general community population, whereas other two studies [32,34] restricted their cohorts to speci c population, such as adult employers.
One study [33] didn't provide gender information. The diversity of included population means potential different educational level, socioeconomic status and mental health level, which might contribute to the bias of synthesized results.

Bias from covariates adjustment
Different studies provided different information of covariates adjustment. All ve studies [29][30][31][32][33][34] adjusted the results for age and two studies [30,33] adjusted the result for gender. One study [32] further adjusted the result for baseline insomnia symptoms, marital status, occupational class, sleep duration and common mental disorders, one study [30] further adjusted for parental education, family income, and parent-reported poor health condition at baseline and one study [29] further adjusted for medical disorder, BMI, physical inactivity, alcohol. One study [31] adjusted for sampling weight.

Bias from de nition of endpoint
Although included studies [29][30][31][32][33][34] all provided clear de nition of insomnia, the de nition varied from studies which might contribute to the bias of synthesized results. Figure 2 showed the ORs of ve studies about smoking and incidence of insomnia. The pooled OR for smoking was 1.07 (95% CI:1.02, 1.13), which indicated a signi cant association between smoking and incidence of insomnia. The heterogeneity was not signi cant (I 2 =14.9%, p=0.310). Funnel plot and Egger's regression revealed that there was no signi cant evidence for publication bias (Egger's regression: p>0.1, Supplementary gure 1). Sensitivity analysis indicated heterogeneity was mainly from one study [31] (Figure 1 and Supplementary gure 2). After reviewing the whole article, we noticed that in this study, participants which had a poor quality of sleep while didn't meet the criterion of insomnia at the baseline were also included in the analysis while in other included studies, this part of participants was excluded.

Smoking and Insomnia Incidence
Besides, PSG (polysomnography) was used to diagnose insomnia in this study, which might nd out potential insomnia that was neglected in other included studies. All these factors might cause the heterogeneity.

Discussion
In this systematic review and meta-analysis, we quantitatively summarized available results from ve cohort studies including about 12.5 thousand people to reveal the substantial association between smoking and incidence of insomnia. The results showed that smoking was signi cantly associated with incidence of insomnia. The heterogeneity was not considerable and was mainly from one study [31].

Association between smoking and incidence of insomnia
Consistent with our result, several cross-sectional studies and epidemiological investigations suggested that smoking might be an independent risk factor for insomnia [19,20,35]. However, the difference in incidence of insomnia between smokers and non-smokers was very small (OR=1.07) which might limit the public health implications of the result. Firstly, since about 30% people might suffer from insomnia [1] and billions of dollars were spent on it [2][3][4], a small improvement of sleep health though anti-smoking could also create considerable public health bene ts. Secondly, it was reported that insomnia was a risk factor of smoking among students [36] and there might be a vicious circle between smoking and insomnia [37], which further highlighted the importance of understanding the relationship between smoking and incidence of insomnia.
For more speci c and quantitative results, we further conducted the strati ed analysis by smoking status. The results showed signi cant association between regular smokers and incidence of insomnia. Several previous studies reported similar association [23,35]. The association between occasional smokers and incidence of insomnia was not signi cant. The main cause of smoking's negative effect on sleep is associated with chemicals in it, so after quitting or decreasing, the impact might reduce. However, one study demonstrated that sleep architecture disturbances among current smokers were not in uenced by the amount of smoking quanti ed by pack-years [38], which indicated there might be other mechanisms of smoking' impact on sleep. Besides, it was reported that occasional smoking was signi cantly associated with excessive daytime sleepiness while regular smoking was not [39]. Firstly, there were only three cohorts in the regular-smoking subgroup and the heterogeneity was minimal. Therefore, we con rmed the result was authentic. There was only one included study in the occasional subgroup which might reduce the credibility of the outcome due to the differences in demographic factors among studies. Secondly, since the effects of smoking were associated with nicotine-acetylcholine receptors in our brain [38], as the daily amount of smoking increased, this kind of receptors might be saturated at some point and the effects of smoking might reach the maximum. The effects of nicotine on our brain were nonspeci c, so while its concentration in the blood reached a certain value, other effects might gradually emerge which could alleviate the effects of nicotine on sleep health. There is still no satis ed explanation based on existing studies. More work on the association between the amount of smoking quanti ed by pack-years and insomnia were needed to be done. To better understand the etiology, concentration of nicotine and other chemicals in cigarette needed to be tested among insomnia population so their effects on sleep could be analyzed.
The pooled analysis didn't indicate a signi cant association in ex-smokers. It was also reported that there were not signi cant differences between former smokers and never smokers for the prevalence of trouble sleeping [40]. Another study showed that longer sleep duration was associated with successful smoking cessation [41]. All these suggested that quitting smoking might be bene cial to sleep health and current smokers might improve sleep quality by quitting smoking.

Subgroup analysis
The region-speci c analysis found signi cant association between smoking and incidence of insomnia in North America and Asia. Epidemiological investigations and cross-sectional studies in these places reported similar conclusion [6,18]. However, the Asian subgroup included only one study and focused on insomnia in adolescents around 13 years old. We needed to be cautious about the result as the age difference of the participants might result in the inhomogeneity. The association was not signi cant in Europe. Consistent with our result, several prior studies [25,42] showed that there was no signi cant association between smoking and insomnia in Europe. Differences in living habits, social atmosphere and gene might contribute to the different results between areas. We noticed that Mediterranean diet, which was popular in Europe, was possibly bene cial factor for sleep health [43]. The pace of life in Europe was much slower than in other areas and lifestyle has been reported to related to insomnia [5]. Besides, genetic diversity might also have an in uence on how human organism reacted to chemicals in cigars [44]. Therefore, it is possible for people to prevent insomnia by changing living habits including diet or slowing down the pace of live.
Subgroup analysis based on age suggested signi cant association both in adults (>18) and juveniles (≤18). Similar results were reported by other studies [7,45,46]. Considering only one included study focused on juveniles, the result might be in uenced by region and smoking status which was associated with the effect of smoking on insomnia as discussed above. More work remained to be done to reveal the association between smoking and insomnia in juveniles. A cohort study of the foetus reported the adverse in uence of in utero exposure to smoking on sleep patterns in preterm neonates [47], which suggested that smoking might in uence sleep health at a very early age. Therefore, people of all ages should try to avoid smoking for improving the sleep health.
Because included studies didn't divide participants into males and females, we couldn't conduct a subgroup analysis based on gender. To explore whether gender was associated with insomnia in smokers, we divided included studies into subgroups by gender ratio (women/men). The result suggested signi cant association in population with gender ratio (women/men) <1, but not in population with gender ratio(women/men) ≥1, that is, smoking has a negative in uence on sleep in population which males are more than females. The point was reported for the rst time. Prior study suggested that females were more likely to suffer from insomnia [24]. After discussion, we concluded that as male smokers were more than female smokers, the sample size of female smokers was not enough in included studies, which might contribute to bias. Wetler et al (1994) [48] suggested that both for males and females, current smoking was associated with several kinds of insomnia symptoms. The association was not signi cant when we changed the endpoints to other insomnia symptoms. Besides, the frequency of insomnia symptoms had an in uence on the association. The pooled result in population with gender ratio (woman/man) ≥1 showed a high heterogeneity and one reported signi cant association between smoking and incident insomnia, the other didn't. As there were only two studies in the subgroup, different region, mental disorder and social status might all contribute to the high heterogeneity, which reduced the credibility of the result. Passive smoking has also been associated with sleep disturbance among pregnant women [49]. To better understand the interaction between gender, smoking and insomnia, more work with female populations and comprehensive analysis of all-cause insomnia incidence remained to be done.

Biological Mechanisms
Our study identi ed smoking, especially regular smoking as a risk factor for insomnia. The speci c biological mechanism of smoking to particulate matters is still evolving. So far, nicotine, the primary addictive component of cigarettes or tobacco, was the most concerned. Nicotine has a known potential for enhancing attention and maintaining a certain level of arousal [50]. These effects are obtained by the central release of dopamine, norepinephrine, serotonin, acetylcholine, all of which have been implicated in the regulation of wakefulness, and by the stimulant effect that nicotine has on cholinergic neurotransmission in the basal forebrain responsible for cortical arousal [50]. In relation to insomnia, nicotine can regulate and change the normal neurotransmitter and disturb sleep architecture both in the early [38] and later stages of the sleep [48]. The changes have been detected by sleep parameters analysis (polysomnography, PSG). Furthermore, nicotine could generate a dose-dependent reduction in sleep e ciency, slow-wave-sleep, REM (rapid eyes movement) sleep and total sleep time [51,52].
Besides, several psychosocial and physiological factors may account for the association between smoking and incident insomnia. Depression and obesity might be underlying factors in the biopsychosocial domain [5]. Cigarette use has been linked to adverse life events and work stress, which are also associated with insomnia symptoms [5]. Future studies are needed to investigate these multiple associations and reveal the biochemical relationship between smoking and insomnia.

Strengths And Limitations
There are several strengths in our study. Firstly, the ve studies we included for meta-analysis and one for systematic review were all cohort studies. We did not include cross-sectional studies, because they couldn't reveal the sequence of smoking and insomnia. Though there were several reviews of this topic, none of them included meta-analysis. So this article was the rst systematic review and meta-analysis in the recent 20 years. Secondly, the quality of included studies were high or moderate, which might contribute to the reliability. Besides, we systematically analyzed the association between smoking status and insomnia, which provided extra basis for the public policy of promoting anti-smoking.
We admitted limitations in the following aspects. Firstly, the number of included studies is not enough, especially for subgroup-analysis. Secondly, the pooled result of smoking and incident insomnia in population with gender ratio (woman/man) ≥1 showed a high heterogeneity. There were two studies in this subgroup and one suggested signi cant association and the other didn't. Excluding any one of them showed different consequences. Thirdly, insomnia includes different kinds of symptoms. We didn't nd enough studies on speci c insomnia symptoms to conduct subgroup analysis while studies focused on different insomnia symptoms might have different results. Besides, insomnia might be a predictive factor for smoking behaviors and we didn't conduct a further discussion due to the limited number of relevant studies. Alcohol, tea, social status, depression and other factors might also have an interaction with smoking and insomnia, which were not discussed in the study. Lastly, there might be grey literature that was not included in this study, leading to potential publication bias.

Study design
We suggest high-quality longitudinal studies and cohort studies with detailed baseline information and regular health examination as ideal study method. Population with different features, such as a speci c job or different age groups and gender composition, should be recorded and synthesized by strati ed analysis. Mental health, geographical region and other potential factors should be taken into account to improve the representativeness of the study.

Smoking status
Future studies should record daily cigarettes that participants smoke and divide them into different groups. The association between smoking and insomnia need more quanti cation.

Adjustment by covariates
During the review of eligible studies, we found various covariates which may exert unstable impact on the summary. We recommend standardized covariates included at least age, sex, BMI, smoking status, alcohol consumption, chronic respiratory diseases, mental disorder, socioeconomic status at an individual level. Further adjustment can be performed by personal habits (such as physical activities, diet) and health status (such as diabetes, family history of disorder). This work requires detailed and elaborate baseline information.

De nition of endpoints
Though all studies provided de nition and diagnostic criteria, some studies lacked preciseness when describing insomnia. We recommend future studies to de ne insomnia according to DSM. And more studies are needed to discuss the relationship between smoking and speci c insomnia symptoms.

Conclusions
Our meta-analysis identi ed smoking as a signi cant risk factor of insomnia. Considering the limited amount of available studies, more high-quality and prospective cohort studies of large sample sizes are needed to understand details of this association. Since both smoking and insomnia are worldwide problems, our study provides new and comprehensive evidence for tobacco control to reduce insomnia.

Declarations
Ethical approval and consent to participate Ethical approval and consent to participate were not necessary for this systematic review.

Consent for publication
Not applicable Availability of data and materials All data analyzed during this study are included in published articles Authors' contributions Nan Hu: design of the work and the analysis of the data; Chunyi Wang: the analysis of the data; Yan Liao: data collection; Qichen Dai: the analysis of the data; Shiyi Cao: design of the work. All authors have read and approved the manuscript.  Table 2 Subgroup-analysis of smoking and incidence of insomnia. I 2 is the value of total variation clarified by heterogeneity. Figure 1 Funnel blot analyzing the publication bias of smoking and insomnia incidence Figure 2 Sensitivity analysis of included studies

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.