Effect of smoking habits on sleep

To evaluate the effect of smoking habits on sleep, data from 1492 adults referred to the Sleep Institute were accessed and divided into 3 categories of smoking status: current, former and non-smokers. Categories of pack-years (<15 and ≥15) defined smoking severity. The association of smoking status and smoking severity with sleep was analyzed for sleep parameters, especially apnea and hypopnea index (AHI) ≥5, more than 5% of total sleep time (TST) spent with oxyhemoglobin saturation (SaO2) <90%, and arousal index. The arousal index was higher among current (21 ± 17) and former smokers (20 ± 17) than non-smokers (17 ± 15; P < 0.04). Former smokers had a higher percent of TST at SaO2 <90% than non-smokers (9 ± 18 vs 6 ± 13; P < 0.04). Former smokers with pack-years ≥15 compared to <15 exhibited higher AHI (22 ± 24 vs 16 ± 21; P < 0.05) and arousal index (22 ± 19 vs 18 ± 15; P < 0.05). Current smokers with pack-years ≥15 compared to <15 exhibited higher arousal index (23 ± 18 vs 18 ± 16; P < 0.05) and percent of TST at SaO2 <90% (11 ± 17 vs 6 ± 13; P < 0.05). Smoking status and pack-years were not associated with AHI ≥5 on logistic regression analysis, but current smokers with pack-years ≥15 were 1.9 times more likely to spend more than 5% of TST at SaO2 <90% than non-smokers (95%CI = 1.21-2.97; P = 0.005). The variability of arousal index was influenced by gender, AHI and current smokers with pack-years ≥15 (all P < 0.01). Smoking habits seem to be associated with arousal and oxyhemoglobin desaturation during sleep, but not with AHI. The effect was more pronounced in current than former smokers.

Wetter et al. (21), in a sleep cohort study, found that current smokers were more closely associated with snoring and moderate or severe sleep-disordered breathing than non-smokers.Kashyap et al. (17), in a controlled study, found that the prevalence of current smokers among patients with an apnea-hypopnea index (AHI) greater than 10 was higher compared with normal controls.Hoffstein (18) demonstrated that, although smoking status and amount of pack-years smoked was not associated with an increased AHI (18), there was a greater number of current smokers and average pack-years among individuals with AHI ≥50 than among those with AHI <10 (18).Finally, Casasola et al. (19) reported that the nocturnal hypoxia Effect of smoking habits on sleep www.bjournal.com.brindex was higher in current smokers than in non-smokers, and that the pack-years smoked index correlated significantly with carboxyhemoglobin concentration as well as the nocturnal hypoxia index (19).These studies notwithstanding, the literature (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21) only partially responds to the question of whether and how smoking habits and sleep are related.Moreover, only a few studies considered the former smoking condition (17,18,20,21) or the number of packyears smoked (17)(18)(19).The purpose of the present study was to evaluate the effect of smoking status (current, former and non-smoker) and pack-years on objective sleep parameters, especially on respiratory parameters.

Patients and Methods
The study was conducted at the Department of Psychobiology at the Universidade Federal de São Paulo (UNIFESP).Data were collected for 3718 database patients referred to the Sleep Institute during the period of September 2004 through February 2005.The inclusion criteria for this data were: adult patients (30 to 70 years of age), not under treatment for obstructive sleep apneahypopnea syndrome (i.e., not in use of positive airway pressure equipment (CPAP/BIPAP) or oral appliances and those who had not been submitted to surgery to treat apnea) and with baseline oxyhemoglobin saturation (SaO 2 ) over 90% (in order to prevent inclusion of patients with primary lung disease).Data for 1492 cases were included in the present investigation.

Database
Anthropometric data included age, gender and body mass index (BMI).Smoking status (current, former and non-smoker) was determined by three possible answers to the question: "Are you a smoker?"("yes", "I was", "no").Moreover, smoking at least 100 cigarettes during life, and with no discontinuation, was used as the criterion for current smokers (22,23).The amount someone had or has smoked, referred as pack-years, was calculated multiplying years smoking by number of packs per day.Periods of smoking cessation were not included in this calculation.Pack-years were grouped into two categories according to severity (<15 and ≥15), according to the cut-off point of the median of the pack-years variable.Daytime sleepiness was evaluated by the Epworth Sleepiness Scale (24) and a score of 10 was utilized as the normality threshold (24).
Nocturnal polysomnograph (PSG) data included in this study were sleep latency, sleep efficiency, percent of sleep stages, arousal index, AHI, baseline SaO 2 and percentage of TST at SaO 2 <90%.PSG recordings were obtained using surface electrodes to continuously record 4-chan-nels of electroencephalography (EEG; C3-A2, C4-A1, O1-A2, O2-A1), electrooculogram (LOG-A2 and ROG-A1), submental and tibialis electromyography, electrocardiography, nasal airflow, thoracic and abdominal respiratory effort, finger oxymetry, and snore microphone.PSG was scored manually according to the criteria of Rechtschaffen and Kales (25) by an experienced sleep technician and reviewed by a sleep medicine physician.EEG arousals lasting more than 3 s were scored manually according to the American Sleep Disorders Association (ASDA) Task Force criteria (26).Apnea and hypopnea are defined according to the American Academy of Sleep Medicine (AASM) Task Force report (27).Apnea is defined as a cessation for at least 10 s (>90% from previous amplitude) of inspiratory airflow.Obstructive apnea is characterized as the absence of airflow with continued respiratory effort.Central apnea is defined as the absence of airflow and respiratory effort.Hypopnea is an airflow reduction (>50% from previous amplitude) lasting 10 s or more and associated with SaO 2 dropping at least 3% and EEG arousal.The AHI is defined as the number of apneas and hypopneas per hour of sleep, and AHI ≥5 was considered to be abnormal.The desaturation index is a decrease greater than 3% of the number of arterial oxygen desaturations per total sleep time (TST).The percent of TST spent with SaO 2 below 90% was also measured, and considered abnormal when above 5%.
This study was based on information in a retrospective database, constructed at the Sleep Institute, which has a capacity of approximately 80 sleep laboratory beds.The records were obtained by internationally recognized instruments [Oxford (England), Sonolab (Brazil), Embla (USA), Polysmith (USA), Stellate (Canada), and Alice 4 (USA)].

Data analysis
All statistical analyses were performed using the SPSS 13.0 statistical software (28) and P < 0.05 was the significance level adopted.Normality was tested by the Shapiro-Wilk test.Data normally distributed were evaluated by the Student t-test or ANOVA followed by the Tukey Honest significant difference test.For non-normal distributed data, differences were analyzed by the Mann-Whitney U-test or Kruskal-Wallis ANOVA followed by the multiple comparison test.Categorical data were analyzed by the chi-square test.
Binary logistic regression was performed to identify possible associations of smoking related variables (status and pack-years) to AHI ≥5 and percent of TST above 5 at SaO 2 <90%.BMI, gender, and age were controlled in both analyses, while AHI only for the second.The non-smokers group was adopted as reference group.Linear regression analyses were conducted in an all sample-based model to S.G.Conway et al.
www.bjournal.com.brdetermine whether gender, age, BMI, AHI, smoking status, and severity could influence arousal index.The same procedure was performed for the Epworth Sleepiness Scale.For each significant predictor, the unstandardized coefficient (B), standard error, and P value were reported.

Results
Tables 1 and 2 summarize the anthropometric and PSG characteristics according to the smoking status and categories of packyears, respectively.The former smokers group was on average the oldest and had higher BMI and Epworth Sleepiness Scale scores compared with non-smokers.Nonsmokers presented the lowest arousal index.Current smokers showed increased N-REM sleep stage 1 and decreased slowwave sleep compared with non-smokers, while former smokers had an increased percent of TST at SaO 2 <90%.
The comparisons of categories of packyears (pack-years ≥15 vs pack-years <15) for former smokers showed that the subgroup of pack-years ≥15 was older, had higher BMI, Epworth Sleepiness Scale score, AHI and arousal index than the subgroup of pack-years <15 (Table 2).The comparisons of categories of pack-years for current smokers showed that the subgroup of pack-years ≥15 was older, had a greater proportion of men, higher arousal index and percent of TST at SaO 2 <90% than the subgroup of pack-years <15 (Table 2).
In a model controlled for BMI, gender and age, the logistic regression analyses showed that neither smoking status (current and former smokers) nor pack-years were associated with AHI ≥5 compared to non-smokers.Also when controlled for BMI, gender, age and AHI, the analysis Table 1.Table 1.Table 1.Table 1 smoking status and pack-year categories were considered, linear regression analysis demonstrated that gender, AHI and current smokers with pack-years >15 influenced the variability of arousal index (Table 4).Considering the same variables for the Epworth Sleepiness Scale, the model revealed an influence of BMI, AHI, current smokers with pack-years ≥15, and former smokers with pack-years ≥15 (Table 5).

Discussion
The present study demonstrates the association of sleepiness, sleep fragmentation, and oxyhemoglobin desaturation during sleep with the amount of pack-years smoked in a clinical population.The effect was more pronounced in current smokers than in former smokers.
Although subjective measurements have suggested that current smokers complain more about difficulty in initiating and maintaining sleep than non-smokers (6,7) and Zhang et al. (20) observed that current smokers exhibited higher sleep latency and lower sleep efficiency in comparison to non-smokers, the present investigation showed no statistical differences in sleep latency and sleep efficiency among current, former and non-smokers.Comparisons of sleep architecture parameters across all status groups showed that current smokers exhibited a higher percent of sleep stage 1 and a lower percent of slow-wave sleep compared to non-smokers, as was reported by Zhang et al. (20).However, the magnitude of these differences has little clinical relevance.The previous findings of subjective nocturnal awakenings (9), difficulty in maintaining sleep (6), difficulty in waking up (7), loss of Table 5.Table 5.Table 5.Table 5 Table 3. Table 3. Table 3. Table 3. Table 3.Odds ratio, confidence intervals and P values from logistic regression model for more than 5% of total sleep time at saturation of oxyhemoglobin less than 90%.
OR 95%CI Odds ratio (OR) was adjusted for body mass index, gender, age, smoking status and apnea-hypopnea index with the non-smokers as the reference group.Log-likelihood ratio: P < 0.001.*P = 0.005 compared to non-smokers (binary logistic regression).
considering the SaO 2 <90% demonstrated that its permanency in more than 5% TST is associated with current smokers with pack-years >15 (OR = 1.9, 95%CI = 1.21 to 2.97; P = 0.005; Table 3).When gender, BMI, age, AHI, sleep quality (6,7), poor sleep (7,8) and poor daytime performance (6) of smokers, are supported by other types of data of the present study, such as: association of smoking habits with sleep fragmentation and with sleepiness.We observed that current and former smokers exhibited higher arousal index compared with non-smokers.These indexes were also higher among current and former smokers with pack-years ≥15 compared with the respective subgroup with pack-years <15.However, the linear regression, considering gender, BMI, age, AHI, smoking status and pack-year categories, showed that, besides gender and AHI, only current smokers with pack-years ≥15 influenced the arousal index value.This suggests that smoking severity plays a role in sleep fragmentation, especially among current smokers.Former smokers presented Epworth sleepiness scores higher than the normal threshold (24), which were also higher compared to non-smokers or among former smokers with pack-years ≥15 compared with those with <15.Although these findings were not observed among current smokers, the linear regression analysis, considering gender, BMI, age, AHI, smoking status and pack-year categories, showed that the Epworth score was influenced by BMI, AHI, current smokers with pack-years ≥15, and former smokers with pack-years ≥15.Once more smoking severity was associated with sleep data.
Reports of the effects of smoking habits on AHI are contradictory (17)(18)(19)21,29).Moreno et al. (29) found an association between smoking and obstructive sleep apnea estimated by the Berlin questionnaire.Wetter et al. (21) reported that current smokers are more closely associated with snoring and moderate or severe levels of AHI than nonsmokers, and this association was aggravated by the number of pack-years.Former smokers after adjustment for confounding factors were not related to snoring (21).Kashyap et al. (17) demonstrated that the percent of non-smokers was greater among patients with AHI <5 than those with AHI ≥10, and that current smokers were 2.5 times more likely to have AHI ≥10 than former and non-smokers combined and that former smokers were not more likely to have AHI ≥10 in comparison to non-smokers.Hoffstein (18) and Casasola et al. (19) reported no significant association between smoking exposure and apnea severity. . . . .Although Casasola et al. (19) suggested the need of a larger sample to assess this possible association, the present study using data from 1492 subjects found no differences in mean AHI among current, former and non-smokers.Moreover, after adjusting for age, BMI and gender, current and former smokers and their packyear subcategories remained not associated with AHI ≥5 compared with non-smokers.
In the present investigation, although former smokers spent more sleep time at SaO 2 <90% than non-smokers, the logistic regression analysis, controlling for BMI, gender, age and AHI, showed that only current smokers with pack-years ≥15 were 1.9 times more likely to spend over 5% of TST at SaO 2 <90% compared with non-smokers.Additionally, current smokers with pack-years ≥15 spent more sleep time at SaO 2 <90% than those with pack-years <15.These results are consistent with the study conducted by Casasola et al. (19), who demonstrated that current smokers had lower oxygen saturation and higher carboxyhemoglobin levels during sleep than non-smokers.They also detected a correlation between number of pack-years and the nocturnal hypoxia index of current smokers.Since smoking habits are not associated with AHI ≥5 but influence arousal index, Epworth score and oxyhemoglobin desaturation during sleep, especially for current smokers, it is possible that smoking habits play a role in upper airway resistance syndrome.Gothe et al. (30) reported that nicotine administration prior to sleep results in a decrease in upper airway resistance, perhaps by an increase in upper airway muscle tone during sleep.However, Kashyap et al. (17) suggested that the progressive decrease in blood nicotine concentrations, after a few hours of sleep, was related to an increase in upper airway resistance, possibly as a result of upper airway inflammatory processes caused by smoking (3,13,19,31).It is possible that former smokers recover from this inflammatory process after a period of cessation of smoking, thus improving arousal index values.Likewise, such behavior should also improve the mechanisms involved in oxyhemoglobin saturation levels during sleep.
Unfortunately, we could not determine the recovery time for sleep respiratory parameters after smoking cessation.Further studies are important to investigate this topic, as well as the dose-response association of smoking and sleeprelated outcomes without the dichotomized pack-years variable.Because non-smokers composed 63.5% of the sample (more than half carrying zero values for pack-years), we dichotomized severity categories according to the median value of pack-years smoked by our group.
Despite the limitations of the present study, our data indicated that, in addition to the influence of smoking status, smoking severity (number of pack-years accumulated during the periods of smoking) plays an independent and important role in impaired sleep parameters, especially those related to sleep-disorder breathing, such as Epworth score, arousal index and oxyhemoglobin desaturation during sleep.These findings were more pronounced in current smokers, supporting the subjective complaints of smokers reported in other studies.Behavioral intervention for smoking cessation should be emphasized because the deleterious effects of smoking on sleep fragmentation and oxyhemoglobin desaturation were decreased in former smokers.Likewise, patients suffering from sleep apnea should be well-informed to avoid smoking in order to prevent greater impairment of sleep parameters.

Table 1 .
. Characteristics of patients according to smoking status.
Data are reported as mean ± SD or as number.BMI = body mass index (kg/m 2 ); AHI = apnea and hypopnea index defined as the number of apneas and hypopneas per hour of sleep; %TST at SaO 2 <90% = percent of total sleep time at saturation of oxyhemoglobin less than 90%.Statistically significant differences within smoking groups: *P < 0.05 (Mann-Whitney U-test); **P < 0.05 (Student t-test); + P < 0.05 (chisquare test).Effect of smoking habits on sleep www.bjournal.com.br

Table 5 .
. Parameter estimates and P values for the regression model with Epworth Sleepiness Scale score as the dependent variable.
AHI = apnea and hypopnea index defined as the number of apneas and hypopneas per hour of sleep.*P < 0.05; **P < 0.01 (linear regression analyses).

Table 4 .Table 4 .Table 4 .Table 4 .Table 4 .
Parameter estimates and P values for the regression model with arousal index as the dependent variable.