Power Spectral Analysis of EEG Activity During Sleep in Cigarette Smokers*

Transcription

1 CHEST Power Spectral Analysis of EEG Activity During Sleep in Cigarette Smokers* Lin Zhang, MD, PhD; Jonathan Samet, MD, MHS; Brian Caffo, PhD; Isaac Bankman, PhD; and Naresh M. Punjabi, MD, PhD, FCCP Original Research SLEEP MEDICINE Background: Research on the effects of cigarette smoking on sleep architecture is limited. The objective of this investigation was to examine differences in sleep EEG between smokers and nonsmokers. Methods: Smokers and nonsmokers who were free of all medical comorbidities were matched on different factors, including age, gender, race, body mass index, and anthropometric measures. Home polysomnography was conducted using a standard recording montage. Sleep architecture was assessed using visual sleep-stage scoring. The discrete fast Fourier transform was used to calculate the EEG power spectrum for the entire night within contiguous 30-s epochs of sleep for the following frequency bandwidths: (0.8 to 4.0 Hz); (4.1 to 8.0 Hz); (8.1 to 13.0 Hz); and (13.1 to 20.0 Hz). Results: Conventional sleep stages were similar between the two groups. However, spectral analysis of the sleep EEG showed that, compared to nonsmokers, smokers had a lower percentage of EEG power in the -bandwidth (59.7% vs 62.6%, respectively; p < 0.04) and higher percentage of EEG power in -bandwidth (15.6% vs 12.5%, respectively; p < 0.001). Differences in the EEG power spectrum between smokers and nonsmokers were greatest in the early part of the sleep period and decreased toward the end. Subjective complaints of lack of restful sleep were also more prevalent in smokers than in nonsmokers (22.5% vs 5.0%, respectively; p < 0.02) and were explained, in part, by the differences in EEG spectral power. Conclusions: Cigarette smokers manifest disturbances in the sleep EEG that are not evident in conventional measures of sleep architecture. Nicotine in cigarette smoke and withdrawal from it during sleep may contribute to these changes and the subjective experience of nonrestorative sleep. (CHEST 2008; 133: ) Key words: cigarette smoking; power spectral analysis; sleep architecture Abbreviations: AHI apnea-hypopnea index; CI confidence interval; REM rapid eye movement; SHHS Sleep Heart Health Study Observational and experimental research has shown that cigarette smoking can adversely affect sleep quality. 1 4 Epidemiologic data 1,2 indicate that complaints of sleep-onset insomnia and lack of restful sleep are more common in smokers than in nonsmokers. Studies 4 using polysomnography have revealed that smokers have numerous alterations in sleep architecture including a decrease in the total sleep time and a shift toward lighter stages of sleep. However, most of the empirical evidence on the adverse effects of smoking on sleep is based on subjective reports instead of objective data, and studies segregating the effects of smoking from those associated with confounding medical conditions are lacking. Conventional measures of sleep architecture are derived from visual inspection of the sleep EEG. While standardized visual sleep scoring is essential for clinical practice, subtle alterations in the sleep EEG cannot be detected with visual scoring. Furthermore, although composite sleep-stage amounts provide a global assessment of sleep quality, they do not offer insight into its temporal evolution. It is certainly possible that smoking has time-dependent effects across the sleep period. During the initial segment of the sleep cycle, smokers may experience difficulty with sleep onset due to the stimulating effects of nicotine. As the night evolves, withdrawal from nicotine may propagate the sleep disturbance. The primary objective of this study was to characterize the sleep EEG in smokers and nonsmokers in the absence of medical conditions. Full-night polysomnography was used to objectively assess sleep architecture CHEST / 133 / 2/ FEBRUARY,

2 in a matched sample of smokers and nonsmokers who were recruited from a community cohort. In addition to comparing conventional sleep stages, techniques of spectral analysis were used to describe differences in the sleep EEG. Study Sample Materials and Methods *From the Departments of Epidemiology (Drs. Zhang, Samet, and Punjabi) and Biostatistics (Dr. Caffo), and Applied Physics Laboratory (Dr. Bankman), Johns Hopkins University, Baltimore, MD. The opinions expressed in this paper are those of the authors and do not necessarily reflect the views of the Indian Health Service. Supported by the National Heart, Lung, and Blood Institute through the following cooperative agreements: HL53940 (University of Washington); HL53941 (Boston University); HL63463 (Case Western Reserve University); HL53937 (Johns Hopkins University); HL53938 (University of Arizona); HL53916 (University of California, Davis); HL53934 (University of Minnesota); HL63429 (Missouri Breaks Research); and HL53931 (New York University). Dr. Punjabi was also supported by grants HL075078, HL086862, and AG The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article. Manuscript received May 18, 2007; revision accepted September 3, Reproduction of this article is prohibited without written permission from the American College of Chest Physicians ( org/misc/reprints.shtml). Correspondence to: Naresh M. Punjabi, MD, PhD, FCCP, Johns Hopkins University Asthma and Allergy Center, 5501 Hopkins Bayview Circle, Baltimore, MD 21224; npunjabi@ jhmi.edu DOI: /chest The current investigation used data from the Sleep Heart Health Study (SHHS), which is a multicenter study on sleepdisordered breathing and cardiovascular disease. Details regarding the design of the SHHS study have been published previously. 5 Briefly, the baseline cohort for the SHHS study was recruited from ongoing epidemiologic studies of cardiovascular and respiratory disease. The SHHS cohort was used to identify a group of matched smokers and nonsmokers using a computerized algorithm. Smokers were those who reported smoking at least 20 cigarettes per day. The selection criteria included the following: age 65 years; and the absence of cardiovascular or respiratory disease, asthma, COPD, diabetes mellitus, and sleep-disordered breathing (apnea-hypopnea index [AHI], 5 events per hour). Participants using any sedative or hypnotic medications were excluded from the study, as were participants with poor-quality sleep studies that precluded visual sleep stage scoring. Finally, the following matching criteria were imposed to minimize confounding: gender; race; age ( 2 years); body mass index ( 1.5 kg/m 2 ); neck circumference ( 4 cm); and AHI ( 1.6 events per hour). Informed consent was obtained from all participants, and the study protocol was approved by the institutional review board of each participating institution. The baseline SHHS cohort consisted of 779 smokers and 2,916 nonsmokers. However, only 40 pairs of smokers and nonsmokers met the strict inclusion criteria outlined above and could be individually matched to each other. Former smokers were not included for two reasons. First, the primary objective of the current study was to characterize the differences in EEG spectral characteristics between smokers and nonsmokers. Second, the requirement that enrolled subjects be matched on several factors and be free of medical comorbidity led to a small number of matched active, former, and never-smokers. Covariate Data Interviewer-administered questionnaires were used to assess prevalent medical comorbidities, smoking history, and caffeine and alcohol consumption. Smoking status was based on selfreports using the following questions: Have you ever smoked cigarettes? and if you smoked before, do you smoke now? Information regarding daily caffeine use and alcohol use 4 h prior to the sleep study was also collected. Given the association between smoking and psychiatric conditions, the Mental Health Inventory-5 6,7 and Mental Component Summary scores 8 derived from the 36-item Medical Outcomes Study short form 9 were used as surrogate measures of mental health. Finally, lack of restful sleep was based on the response to the question How often do you feel unrested during the day, no matter how many hours of sleep you had? Polysomnography An unattended polysomnogram was conducted at home and consisted of continuous recordings of the following channels: C 3 -A 2 and C 4 -A 1 EEG; right and left electrooculograms; an ECG; a chin electromyogram; oxyhemoglobin saturation; inductance plethysmography; oronasal thermocouple; and body position. Details of polysomnographic equipment, hook-up procedures, failure rates, scoring, and quality assurance and control have been previously described. 10 Sleep stage scoring and respiratory analysis was performed by trained technicians according to current guidelines. 11 Spectral Analysis of the EEG The C 3 -A 2 and C 4 -A 1 EEG recordings were sampled at 125 Hz and analyzed using the techniques of the discrete fast Fourier transform. The fast Fourier transform was conducted on an EEG record length of 5-s to obtain a frequency resolution of 0.2 Hz. Each 5-s segment of the EEG signal was first windowed with a Hanning tapering window prior to computing the power spectra. 12 The power content (expressed as V 2 ) for each 30-s epoch of sleep was determined as the average power across the six 5-s segments of the EEG. The spectral distribution was categorized into the following frequency bands: (0.8 to 4.0 Hz); (4.1 to 8.0 Hz); (8.1 to 13.0 Hz); and (13.1 to 20.0 Hz). The -frequency bandwidth was further subdivided into 1 (13.1 to 16.0 Hz) and 2 (16.1 to 20.0 Hz). The power in each frequency bandwidth was expressed as a percentage of total power in each 30-s epoch of sleep. Frequencies 0.8 Hz were excluded to remove the effects of low-frequency artifacts (eg, sweating and respiration). Synchronizing the epoch-by-epoch spectral estimation with visually assessed sleep stages allowed for restricting the analysis only to sleep epochs and thus permitted the removal of movement time or wakefulness. Statistical Analysis The power in each spectral bandwidth and the lack of restful sleep were modeled as a function of smoking status. First, the power in each spectral bandwidth was compared between smokers and nonsmokers for the entire night with the paired t test. Subsequently, the temporal profile of relative power within the different frequency bands was examined. To parameterize the 428 Original Research

3 time-dependent evolution of EEG power spectra as a function of smoking status, the methods of fixed-knots linear cubic regression splines 13 were implemented to allow for a flexible depiction of temporal trends. Multivariable logistic regression models were used to determine whether subjective reports of lack of restful sleep were explained by differences between smokers and nonsmokers in EEG power spectra. All analyses were conducted using a statistical software package (SAS, version 9; SAS Institute, Inc; Cary, NC). Results Forty smokers and nonsmokers met the strict exclusion criteria, and were individually matched on age, gender, race, body mass index, neck circumference, and AHI (Table 1). Smokers reported smoking an average of 25.3 cigarettes per day (range, 20 to 50 cigarettes per day). Lack of restful sleep was reported by 5.0% of the nonsmokers and 22.5% of the smokers (p 0.02). Measures of mental health status were similar between smokers and nonsmokers as was the amount of self-reported alcohol consumption. However, a higher proportion of smokers reported daily consumption of caffeinated products (Table 1). Nevertheless, caffeine consumption was not associated with the results of the EEG power spectral analysis or self-report of lack of restful sleep. Conventional sleep-stage statistics were similar between the two groups (Table 2). Table 3 shows the average spectral power from the entire night by smoking status. Compared to nonsmokers, smokers on average had a higher percentage of -power (15.6% vs 12.5%, respectively; p 0.001) and a lower percentage of -power (59.7% vs 62.6%, respectively; p 0.04) during sleep. However, no differences were noted in the percentages of -power or -power (Table 3). The absolute difference of 2.9% and 3.1%, respectively, in -power and -power between smokers and nonsmokers should be interpreted within the context of the observed maximum-to-minimum change in power for each bandwidth during sleep. In nonsmokers, EEG -power is at a minimum at sleep onset and constitutes approximately 53.8% of a sleep epoch (Fig 1, bottom). It reaches a peak of 73.9% during the first third of the sleep period, resulting in an overall maximum-tominimum change of 20.1%. Similarly, -power at sleep onset is at a maximum of approximately 16.4% in nonsmokers and reaches a nadir value of 8.7% (Fig 1, top). Thus, the resulting maximum-to-minimum change in EEG -power in nonsmokers is 7.8%. Using these maximum-to-minimum changes in EEG spectral power from nonsmokers ( -power, 20.1%; -power, 7.8%) as references, the previously observed absolute differences in -power and for -power between smokers and nonsmokers would then amount to a proportional difference of 39.8% (95% CI, 16.7 to 63.1) and 14.6% (95% CI, 0.7 to 28.4), respectively. Temporal analyses of the sleep EEG showed distinct patterns between the two groups with the largest differences in EEG -power and -power. As shown in Figure 1, top, an initial absolute difference of 4.3% (95% CI, 3.8 to 4.8) in -power was noted between smokers and nonsmokers. This difference, however, decreased toward the end of the sleep period with a significant interaction between smoking status and time since sleep onset (p ). Differences were also noted in the temporal evolution of -power across the night (Fig 1, bottom). -power was greater initially in nonsmokers by 4.7% (95% CI, 3.7 to 5.7) compared to smokers. The difference also diminished toward the end of the sleep period (p for the interaction with time). EEG -power and -power during sleep did not show time-varying differences (data not shown). Further analyses revealed that the temporal variations in the EEG power spectra were distinct during periods of non-rapid eye movement (REM) and REM sleep. The previously observed differences in -power and -power were confirmed during non-rem sleep. Smokers, on average, had Table 1 Characteristics of Smokers and Nonsmokers* Characteristics Smokers Nonsmokers p Value Age, yr Male gender 20 (50.0) 20 (50.0) 1.00 White race 36 (90.0) 36 (90.0) 1.00 Body mass index, kg/m Neck circumference, cm Daily caffeine use 38 (95.0) 30 (75.0) 0.03 Cigarette use 38 (95.0) 0 (0.0) Alcohol use 6 (15.0) 5 (12.5) 0.25 Mental Health Index-5 score Mental component summary score *Values are given as the mean SD or No. (%), unless otherwise indicated. Daily caffeine use positive response to the daily consumption of caffeinated tea, coffee, or sodas. Group differences by smoking status were determined by the Fisher exact test for categoric variables and paired t test for continuous variables. Cigarette and alcohol use 4 h before the sleep study. CHEST / 133 / 2/ FEBRUARY,

4 Table 2 Parameters of Conventional Sleep-Stage Data in Smokers and Nonsmokers* Sleep Architecture Smokers Nonsmokers p Value Total sleep time, h Sleep latency, min 22.3 ( ) 18.5 ( ) 0.3 Stage 1 sleep, % total sleep time 4.8 ( ) 4.8 ( ) 0.8 Stage 2 sleep, % total sleep time Slow-wave sleep, % total sleep time 15.8 ( ) 16.9 ( ) 0.4 REM sleep, % total sleep time REM latency, min AHI, events/h *Values are given as the mean SD or median (25th to 75th percentiles) for non-normally distributed variables. Group differences by smoking status were determined by paired t tests. Denotes the latency to the first onset of sleep and first episode of REM sleep, respectively. 3.7% (95% CI, 1.0 to 6.4) less -power and 3.4% (95% CI, 1.6 to 5.1) more -power in non-rem sleep than nonsmokers. In contrast, group differences in EEG power spectra during REM sleep were only noted in the -bandwidth. Smokers had 3.2% (95% CI, 1.1 to 5.3) more -power than nonsmokers in REM sleep. To assess whether the differences in -power and -power could explain the higher prevalence of self-reported lack of restful sleep in smokers, multivariable logistic models were constructed. Modeling building was based on a stepwise inclusion of spectral parameters that were different between the two matched groups. Inclusion of -power in these multivariable models did not lead to any significant change in the odds ratio of association between smoking status and lack of restful sleep. Thus, for the sake of parsimony, the amounts -power and -power were retained as covariates in the final model. After adjustments for the average overall amount of -power and -power, the prevalence odds ratio for lack of restful sleep associated with smoking decreased from 5.52 (95% CI, 1.01 to 27.43) to 4.1 (95% CI, 0.74 to 23.0), suggesting that differences in EEG spectral power were associated with the higher prevalence of self-reported lack of restful sleep in smokers. Table 3 Results of Full-Night EEG Power Spectral Analysis in Smokers and Nonsmokers* EEG Frequency Smokers, % Nonsmokers, % p Value ( Hz) ( Hz) ( Hz) ( Hz) ( Hz) ( Hz) *Values are reported as mean SD, unless otherwise indicated. Results of power spectral analyses are summarized as the percentage of power within a specified frequency range as a proportion to the total power in the respective epoch of sleep. Determined by paired t tests. Discussion The last few decades have seen substantial advancements in our knowledge of the harmful health EEG α power (%) EEG δ power (%) 22% 20% 18% 16% 14% 12% 10% 8% Sleep Onset 75% 70% 65% 60% 55% 50% Smoker Smoker Nonsmoker p for interaction < Time after sleep onset (hours) Nonsmoker p for interaction: Sleep Onset Time after sleep onset (hours) Figure 1. Temporal course of EEG -power (top) and -power (bottom) during sleep by smoking status. EEG-power spectral curves were computed using flexible linear regression spline techniques (see Materials and Methods section). 430 Original Research

5 effects of cigarette smoking. The present study shows that cigarette smoking can alter sleep architecture independent of factors such as age, gender, race, anthropometric measures, caffeine and alcohol consumption, medical comorbidity, and mental health status. Despite similar sleep stage architecture, the EEG power spectrum in smokers was shifted toward higher frequencies compared to nonsmokers. Differences in the sleep EEG between smokers and nonsmokers were time dependent, with the largest dissimilarity occurring in the early part of the sleep period. Although the absolute group differences may appear small, these have to be viewed against the background of the maximal possible change in EEG power in any given frequency band over the course of the night. Finally, the higher prevalence of self-reported unrestful sleep in smokers was explained, in part, by the differences in the EEG spectral power. Experimental data from animal models 14,15 and nonsmoking healthy subjects 16,17 indicate that nicotine administration can alter normal sleep architecture. Acute exposure to nicotine in healthy volunteers can have alerting effects, and a dose-dependent decline in slow-wave sleep, REM sleep, and total sleep time. 16,17 These effects arise because nicotine stimulates the central release of dopamine, norepinephrine, serotonin, and acetylcholine, all of which have been implicated in the regulation of wakefulness. 18 Although causal mechanisms cannot be directly inferred from the current study, characterizing the overnight evolution of the EEG provides a better understanding of the pathophysiologic effects of nicotine on sleep. It is well established that blood nicotine levels are highest at bedtime in habitual smokers. 19 Thus, differences in sleep architecture between smokers and nonsmokers would be greatest in the early part of the sleep period. In fact, the shift in the EEG power spectrum toward higher frequencies in smokers lends support to the notion that the stimulant effects of nicotine may have an essential role, at least initially, in altering sleep. Given that the half-life of nicotine is approximately 2 h, minor withdrawal from nicotine throughout the night could further impact sleep architecture in the latter part of the sleep period. Alterations in sleep architecture, as noted herein, are of clinical significance as they may explicate the range of subjective sleep-related symptoms associated with smoking, including nonrestorative sleep. Several strengths and limitations of the current study merit discussion. First, the exclusion of prevalent cardiovascular or pulmonary disease and other medical conditions allowed for an unconfounded examination of the adverse effects of smoking on sleep architecture. Second, rigorous matching of smokers and nonsmokers minimized the potential influence of confounding factors on nighttime EEG patterns. Third, the comparability of mental health status between smokers and nonsmokers diminishes the possibility that mental health conditions are responsible for the observed differences between the two groups. Fourth, in contrast to previous work, 20 the use of a community-based sample with home polysomnography enhances the generalizability of the reported findings. A limitation of this study is the modest sample size, which precluded the examination of a dose-response relationship between the amount smoked and EEG spectral power. However, previous analyses 4 have shown that conventional sleep stage amounts are also not associated with the number of pack-years smoked. Another limitation is that smoking status was based on self-reports without corroborating biochemical measurements. Comparisons of self-reported and biochemically assessed smoking status 21,22 have found that the discrepancies in measurement are not large. Restricting the analysis to those participants without a medical comorbidity and high-quality EEG records may have omitted the most susceptible subjects with disrupted sleep (ie, smokers with chronic medical conditions and worse sleep quality). Such exclusion would minimize the associations of interest, and thus the reported differences between smokers and nonsmokers are, at best, conservative in magnitude. Finally, while subjects with sleep-disordered breathing were excluded from the study, residual confounding due to the inclusion of those subjects with upper airway resistance syndrome, which is known to be associated with increased -frequency power during sleep, 23 is certainly possible. In conclusion, the current study shows that smoking is associated with alterations in the sleep EEG. Finding similar sleep architecture in smokers and nonsmokers highlights the informativeness of quantitative methods that fully utilize the temporal detail of the sleep EEG. The present study also shows that smokers report more subjective sleep disturbance, which may be a due to differences in the sleep EEG. Although the exact mechanisms underlying the subjective and objective sleep disturbance in smokers need to be further defined, exposure to and withdrawal from nicotine likely represent important causal factors. Finally, an important implication of the current study is that spectral analysis of the sleep EEG can uncover differences that may not be evident with traditional sleep stage measures. Thus, future research efforts assessing the adverse effects of a specific factor or medical condition on sleep should include the use of quantitative EEG analysis. CHEST / 133 / 2/ FEBRUARY,

6 References 1 Wetter DW, Young TB. The relation between cigarette smoking and sleep disturbance. Prev Med 1994; 23: Phillips BA, Danner FJ. Cigarette smoking and sleep disturbance. Arch Intern Med 1995; 155: Wetter DW, Fiore MC, Baker TB, et al. Tobacco withdrawal and nicotine replacement influence objective measures of sleep. J Consult Clin Psychol 1995; 63: Zhang L, Samet J, Caffo B, et al. Cigarette smoking and nocturnal sleep architecture. Am J Epidemiol 2006; 164: Quan SF, Howard BV, Iber C, et al. The Sleep Heart Health Study: design, rationale, and methods. Sleep 1997; 20: Berwick DM, Murphy JM, Goldman PA, et al. Performance of a five-item mental health screening test. Med Care 1991; 29: Rumpf HJ, Meyer C, Hapke U, et al. Screening for mental health: validity of the MHI-5 using DSM-IV axis I psychiatric disorders as gold standard. Psychiatry Res 2001; 105: Beusterien KM, Steinwald B, Ware JE Jr. Usefulness of the SF-36 Health Survey in measuring health outcomes in the depressed elderly. J Geriatr Psychiatry Neurol 1996; 9: Ware JE Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36): I. Conceptual framework and item selection. Med Care 1992; 30: Redline S, Sanders MH, Lind BK, et al. Methods for obtaining and analyzing unattended polysomnography data for a multicenter study: Sleep Heart Health Research Group. Sleep 1998; 21: Rechtschaffen A, Kales A. Manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Washington, DC: US Government Printing Office, 1968; NIH Publication No Bracewell RN. Sampling and series: the Fourier transform and its applications. New York, NY: McGraw-Hill, 2000: de Boor C. A practical guide to splines. New York, NY: Springer, Salín-Pascual RJ, de la Fuente JR, Galicia-Polo L, et al. Effects of transderman nicotine on mood and sleep in nonsmoking major depressed patients. Psychopharmacology (Berl) 1995; 121: Salín-Pascual RJ, Moro-Lopez ML, Gonzalez-Sanchez H, et al. Changes in sleep after acute and repeated administration of nicotine in the rat. Psychopharmacology (Berl) 1999; 145: Davila DG, Hurt RD, Offord KP, et al. Acute effects of transdermal nicotine on sleep architecture, snoring, and sleep-disordered breathing in nonsmokers. Am J Respir Crit Care Med 1994; 150: Gillin JC, Lardon M, Ruiz C, et al. Dose-dependent effects of transdermal nicotine on early morning awakening and rapid eye movement sleep time in nonsmoking normal volunteers. J Clin Psychopharmacol 1994; 14: Kenny PJ, Markou A. Neurobiology of the nicotine withdrawal syndrome. Pharmacol Biochem Behav 2001; 70: Benowitz NL, Kuyt F, Jacob P III. Circadian blood nicotine concentrations during cigarette smoking. Clin Pharmacol Ther 1982; 32: Soldatos CR, Kales JD, Scharf MB, et al. Cigarette smoking associated with sleep difficulty. Science 1980; 207: Patrick DL, Cheadle A, Thompson DC, et al. The validity of self-reported smoking: a review and meta-analysis. Am J Public Health 1994; 84: Murray RP, Connett JE, Istvan JA, et al. Relations of cotinine and carbon monoxide to self-reported smoking in a cohort of smokers and ex-smokers followed over 5 years. Nicotine Tob Res 2002; 4: Guilleminault C, Do KY, Chowdhuri S, et al. Sleep and daytime sleepiness in upper airway resistance syndrome compared to obstructive sleep apnoea syndrome. Eur Respir J 2001; 17: Original Research