An automatic ambulatory device for detection of AASM defined arousals from sleep: the WP100

Sleep Medicine 4 (2003) 207–212 www.elsevier.com/locate/sleep Original article An automatic ambulatory device for detection of AASM defined arousals from sleep: the WP100 Giora Pillara, Amir Barb, Michal Betitob, Robert P. Schnallb, Itsik Dvirb, Jacob Sheffyb, Peretz Lavie*,a,1 a Sleep Laboratory, Bruce Rappaport Faculty of Medicine, Technion – Israel Institute of Technology, Haifa, Israel b Itamar Medical Ltd., Caesarea, Israel Received 16 July 2002; received in revised form 1 October 2002; accepted 23 October 2002 Abstract Objectives and background: Arousals from sleep are associated with increased sympathetic activation and therefore with peripheral vasoconstriction. Sleep fragmentation in the form of multiple arousals is associated with daytime somnolence and cognitive impairment; however, manual scoring of arousal is time consuming and problematic due to relatively high inter-scorer variability. We have recently shown that automated analysis of in-lab recorded peripheral arterial tone (PAT) signal and the pulse rate derived from it can accurately assess arousals from sleep as defined by the American Academy of Sleep Medicine (AASM). In the current study we sought to extend these findings to the Watch_PAT100 (WP100), an ambulatory device measuring PAT, oximetry and actigraphy. Methods: Sixty-eight subjects (61 patients referred to the sleep lab with suspected obstructive sleep apnea and seven healthy volunteers, mean age 46.3 ^ 14.2 years) underwent a whole night polysomnography (PSG) with simultaneous recording of PAT signal by the ambulatory WP100 device. The PSG recordings were blindly manually analyzed for arousals based on AASM criteria, while PAT was scored automatically based on the algorithm developed previously. Results: There was a significant correlation between AASM arousals derived from the PSG and PAT autonomic arousals derived from the WP100 (R ¼ 0:87, P , 0:001), with a good agreement across a wide range of values. The sensitivity and specificity of PAT in detecting patients with at least 20 arousals per hour of sleep were 0.80 and 0.79, respectively, with a receiver operating characteristic curve having an area under the curve of 0.87. Conclusions: We conclude that automatic analysis of peripheral arterial tonometry signal derived from the ambulatory device Watch_PAT100 can accurately identify arousals from sleep in a simple and time saving fashion. q 2003 Elsevier Science B.V. All rights reserved. Keywords: Sleep; Autonomic nervous system; Arousals; Peripheral arterial tone; Ambulatory monitoring; Sympathetic activation 1. Introduction Sleep fragmentation in patients with sleep apnea syndrome can result in non-restorative sleep and consequent daytime sleepiness and impairment of cognitive and psychomotor performance [1 – 3]. Even normal subjects become sleepier and their mood is impaired during the day following experimental sleep fragmentation by brief arousals [4]. Thus, the number of arousals is a useful * Corresponding author. Giora Pillar, MD, Phd, Sleep Laboratory, Gutwirth Building, Technion City, Haifa 32000, Israel. Fax: þ 972-48537404. E-mail address: gpillar@tx.technion.ac.il (G. Pillar). 1 Amir Bar, Michal Betito, Robert Schnall and Jacob Sheffy are employees of Itamar Medical and Giora Pillar and Peretz Lavie are consultants for Itamar Medical Ltd. marker of sleep quality, independent of traditional sleep quality markers such as sleep latency, wake after sleep onset and sleep efficiency. The currently recommended criteria for scoring arousals consist of a notable EEG shift for at least 3 s but no more than 15 s during all NREM stages of sleep, assuming sleep is recorded prior to and following the event for at least 10 s. Since EEG alpha waves or mixed frequency waves are common during REM sleep, the definition of arousal during REM sleep relies on a combination of EEG defined arousal and increased EMG or body movements [5]. These criteria are rather difficult to determine, and a relatively large inter-scorer variability has been reported in scoring arousals from sleep [6,7]. Thus, an automatic and reliable method to detect arousals has been sought [8]. We have recently reported that an automatic analysis of peripheral arterial tone (PAT) signal recording – 1389-9457/03/$ - see front matter q 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S1389-9457(02)00254-X 208 G. Pillar et al. / Sleep Medicine 4 (2003) 207–212 a simple, reproducible and time saving procedure – can accurately detect arousals from sleep [9]. In this previous study, PAT signal has been recorded as an additional channel in a standard polysomnography (PSG) set-up, with sleep/wake scoring derived from the PSG. The fact that standard PSG is a relatively cumbersome and expensive procedure drives researchers to develop ambulatory methods and devices. For the diagnosis of obstructive sleep apnea (OSA), several devices have been produced [10 –13], yet none have gained enough popularity to be widely used for clinical purposes. For the detection of sleep fragmentation, the pulse transit time (PTT) method has been introduced. Pitson et al. reported a good correlation between PTT and EEG frequency shifts in response to external stimuli in normal subjects [14]. PTT could also, to some extent, detect sleep disordered breathing events [15]. Argod et al. found a reasonable agreement between standard scoring of PTT in detecting non-apneic obstructive respiratory events, but reported a very high inter-observer variability in the scoring of both (30 – 37%) [16]. In the current study, we sought to examine and validate the accuracy of the recently developed ambulatory WP100 device (Watch_PAT100) in the detection of arousals from sleep, as defined by the American Academy of Sleep Medicine (AASM) [5]. 2. Methods 2.1. Subjects The study group consisted of 61 consecutive adult patients referred to the Technion Sleep Disorders Center for evaluation of presumed obstructive sleep apnea syndrome (OSAS) and an additional seven young healthy volunteers, recruited via advertisements in the Faculty of Medicine, with no complaints of sleep disruption, daytime sleepiness or snoring. The healthy volunteers were free of disease and medications. The exclusion criteria for the suspected OSAS patients were: permanent pacemaker, non-sinus cardiac arrhythmias, peripheral vasculopathy or neuropathy, severe lung disease, S/P bilateral cervical or thoracic sympathectomy, finger deformity that precludes adequate sensor application, use of alpha-adrenergic receptor blockers (24 h washout period required), and alcohol or drug abuse during the last 3 years. The study was approved by the Rambam Medical Center Committee for Studies in Human Subjects, and patients signed an informed consent prior to participation. Fifty-four of the participants were males and 14 were females. A wide range of OSAS severities were represented in the study group, with the respiratory disturbance index (RDI) ranging from one to 118 events/phs. Twenty percent of the subjects had hypertension and 4% had coronary artery disease. 2.2. Protocol All participants underwent a whole night PSG (Embla system, Flaga HF, Iceland) with simultaneous recordings of the WP100 device (Itamar Medical Ltd., Caesarea, Israel). Prior to bedtime patients completed a sleep questionnaire including physical data (e.g. weight and height), general health condition and medical history, medication usage, sleep habits, and the Epworth Sleepiness Scale (ESS) [17]. Lights-off was no later than midnight, and lights-on was at 06:00 h. 2.3. PSG Overnight PSG was performed according to standard laboratory protocol, using computerized PSG with the following channels: two EEG (C3-A2 and O2-A1), EOG, submental EMG, arterial oxygen saturation, nasal – oral airflow (thermistors and nasal pressure), EKG, chest and abdominal wall motion (piezo electrodes), bilateral anterior tibialis EMG, and body position. Sleep was staged according to standard criteria [18]. Arousals were defined according to the AASM guidelines [5]. An EEG frequency shift of at least 3 s but no more than 15 s during non-REM sleep was scored as an arousal; during REM sleep an increase in EMG was required as well, and in both cases at least 10 s of sleep prior to and following the event was required. The arousal index (ARI) was calculated by dividing the total number of arousals by the number of hours of sleep, and has been termed PSG-ARI. Respiratory events were scored according to the AASM guidelines for measurement in clinical research [19]. The RDI was calculated as the number of apneas plus hypopneas divided by the number of hours of sleep. 2.4. The Watch_PAT 100 device The WP100 device is an ambulatory system comprised of a battery powered consol unit, mounted just above the wrist, with embedded actigraph and oximetry. It has two finger mounted probes: pulse oximetry and PAT. This system has been described in detail elsewhere [20]. In brief, it records three signals (actigraphy, PAT, oximetry) with a sampling rate of 100 Hz, and stores the data throughout the sleep study on a removable flash disc. A fourth channel, pulse rate, is derived from the PAT signal. Sleep/wake is determined by the actigraphy, and the arousals are scored automatically using an improvement of an algorithm, previously described [9]. The algorithm scores an arousal if one of the following conditions is fulfilled within an epoch scored as sleep: 1. An association between two events – an attenuation of the PAT signal amplitude (vasoconstriction) below a threshold and an increase in pulse rate above a threshold. The thresholds for both events are adaptive and set in a 209 G. Pillar et al. / Sleep Medicine 4 (2003) 207–212 first pass learning process of the patient’s recorded data file according to the ratio between the number of these events. 2. An association between PAT signal amplitude attenuation of more than 40% and short movements of the patient, detected by the energy of the actigraphic signal. The total number of arousals scored was divided by the number of hours of sleep (as assessed by the WP100), and termed PAT-AAI. 2.5. Data and statistical analysis Different personnel separately and blindly performed the PSG manual scoring and the PAT automatic scoring. The primary outcome measures in this study were the AASMbased arousal indices (PSG-ARI), which were considered the gold standard, and the PAT-based autonomic arousal indices (PAT-AAI), which were the evaluated measures. The agreement between these two measures was assessed in three ways. First, correlation analysis was performed. Second, Bland – Altman analysis was carried out to assess potential range-dependant agreement. Finally, in order to evaluate the efficacy of the PAT automatic analysis as a potential tool for diagnosing a sleep arousal disorder, a threshold of AAI ¼ 20 [21] for abnormality was defined, a receiver operating characteristic (ROC) curve plotted, and its area under the curve (AUC) calculated. This curve joins points on an X_Y plain (x ¼ 1 2 specificity, y ¼ sensitivity) for all possible values of the PAT-AAI thresholds. The AUC is considered a measure of the overall efficacy of the score; an AUC value of 0.5 indicates a non-significant score for separating normal from abnormal patients and a value close to 1.0 indicates a very efficient score. In addition, the correlation of both PSG-ARI and PATAAI with the Epworth Sleepiness Score was determined to assess their ability to predict subjective sleepiness. Table 1 Characteristics of the study group RDI n Age (years) BMI ESS ,10 10–20 21–40 .40 Overall 14 35 ^ 16 25 ^ 4 6^6 14 44 ^ 14 27 ^ 4 9^5 15 50 ^ 9 27 ^ 5 10 ^ 5 25 52 ^ 12 32 ^ 6 12 ^ 6 68 46 ^ 14 28 ^ 6 9.5 ^ 6 RDI, respiratory disturbance index (events/hour of sleep); n, number of subjects; BMI, body mass index (kg/m2); ESS, Epworth Sleepiness Scale (0–24). Data are presented as the mean ^ SD. good agreement between the two parameters across a wide range of arousal indices. Fig. 3 shows the ROC curve for PAT identification of patients with pathological AAI, as defined by a threshold of 20 arousals per hour of sleep [21] when taking the AASM-based scoring as ‘gold standard’, with an AUC of 0.87. The sensitivity and specificity of the WP100 in detecting patients with at least 20 arousals/phs were 0.80 and 0.79, respectively. Finally, although statistically significant, there were only poor correlations between the arousal indices (either by standard criteria or by PAT) and subjective sleepiness as assessed by ESS (R ¼ 0:33 and R ¼ 0:34, respectively, P , 0:001 for both). 4. Discussion This study shows that the standard AASM-based ‘EEG arousals’ can be accurately assessed by measuring ‘autonomic arousals’ using the WP100 ambulatory device. This is consistent with the observation that arousals are 3. Results Characteristics of the study population are presented in Table 1. The average age and BMI were 46 ^ 14 years and 28 ^ 6 kg/m2, respectively. The average RDI for the whole group was 34 ^ 26 events/h. As can be seen, there were similar numbers of subjects in the various apnea severity ranges. The average ESS score for the whole group was 9.5 ^ 6, with the score tending to be higher as the severity of apnea increased. Fig. 1 displays a scatter graph of the PSG-ARI (gold standard) vs. the PAT-AAI values for the whole study population with the calculated correlation coefficient. There was a good and statistically significant correlation between the two parameters (R ¼ 0:87, P , 0:0001). Fig. 2 displays the Bland –Altman plot for PSG-ARI and the PAT-AAI for the study population. There was a Fig. 1. PSG-ARI vs. PAT-AAI. A very high and statistically significant correlation (R ¼ 0:87, P , 0:0001) was found between the PAT-AAI (PAT-based autonomic arousal index) and the PSG-ARI (arousal index derived from the PSG based on AASM criteria). 210 G. Pillar et al. / Sleep Medicine 4 (2003) 207–212 Fig. 2. Bland–Altman presentation of ASDA-ARI vs. PAT-ARI. Across a wide range of arousals frequencies, there was a good agreement between PAT-ARI (PAT-based arousal index) and the PSG-ARI (arousal index based on AASM criteria). associated with sympathetic activation, and can be accurately measured by an in-lab PAT channel added to a standard PSG [9]. Until now, ambulatory sleep monitoring equipment has been focused on measuring only the sleep/wake state (actigraphy) [22], sleep apnea indices (primarily oximetry, but also other methods) [11,13,23,24], or multi-channel measurement (home PSG) [10,12,25]. In the current study we have concentrated on measuring autonomic arousals from sleep, and have used a relatively novel channel (PAT signal) monitored by an ambulatory device (WP100). Brief arousals from sleep may impair cognitive and psychomotor performance, particularly in sleep-related breathing disorders [1,4]. Thus, quantification of arousals may have an important clinical role in patient assessment. However, EEG arousals, as currently defined, have several disadvantages. EEG recordings must be done either in the lab or by a relatively cumbersome ambulatory monitoring system, and scores are unreliable, primarily due to a considerable inter-scorer variability. Drinnan et al., using experts from 14 sleep laboratories to evaluate the reprodu- Fig. 3. ROC curve for identifying pathological arousals from sleep (threshold: 20 arousals per hour of sleep) based on PAT vs. standard criteria. The AUC is 0.87, yielding potentially high sensitivity and specificity in diagnosing pathological arousal frequency by PAT, when the gold standard is the AASM criteria based on PSG. cibility of EEG arousal scores as defined by the AASM, report a rather large disagreement and considerable variability in the scoring of these events [7]. In another study Loredo et al., evaluating various types of arousals, showed good inter-scorer reproducibility in scoring Periodic Leg Movements (PLMs) or respiratory events accompanied by arousals characterized by increased EMG. Poor reproducibility was demonstrated in the scoring of ‘classic’ AASM defined arousals [6]. Thus, an automated method of detecting sleep fragmentation, which by definition does not present an inter-scorer variability, seems warranted. Two additional methods based on sympathetic activation have been previously attempted. The PTT has initially demonstrated some promising results, and also has the advantage of potential use in the home environment [26]. However, it has not been compared in large-scale studies to standard measures, and has been shown to have the disadvantage of high inter-observer variability [16]. Davies et al. used arousal combined with inspiratory blood pressure changes as a potential marker for disturbed sleep and disordered breathing in sleep [27]. Although blood pressure can be measured non-invasively by an ambulatory device, and the blood pressure profile seems to be useful in identifying patients with OSA, this method has not been further investigated – perhaps because of the impracticality of continuous blood pressure measurements throughout the night. The monitoring of PAT, which we have used in the current study, has several advantages. First, the very simple device is located only on the hand and requires minimal technical intervention in patient preparations and recording. The applied pressure to the finger (approximately 50 mmHg) did not cause any adverse effect or discomfort to the subjects. Second, the analysis of the PAT data is automatic (computerized), which makes it reproducible, objective and time saving. In the current study we have not quantified time consumption by each method of arousal assessment, but we roughly estimate that AASM-based PSG arousal scoring took some 40 –60 min per record, while the automatic PAT-based arousal count, with all overheads (such as downloading the patient file to a work station), takes less than 5 min per record. Third, the incorporation of actigraphy enables this automatic analysis to be applied during sleep periods only. Although the aim of our study was to assess the ability of the ambulatory device WP100 to detect arousals from sleep, in the present study we have used this device in the sleep lab in order to compare the measured arousals to standard measurements by full PSG (AASM criteria). All three methods of comparisons between the AASM and PATbased AAI (correlation analysis, Bland – Altman analysis and ROC) revealed reasonably good agreement, suggesting that the PAT-based algorithm may accurately reflect sleep fragmentation. As discussed elsewhere [9], PAT-based arousals are recognized based on analysis of two important sympathetic-related parameters: digital vasoconstriction and pulse rate changes, both measured by the finger G. Pillar et al. / Sleep Medicine 4 (2003) 207–212 probe. It has been shown that arousals from sleep are associated with increased sympathetic activation [28 – 32]. The WP100e algorithm, which detects autonomic arousals, is activated only in epochs of sleep, as determined by the actigraphic part of the WP100. In this study, totally independent analysis of the PSG and WP100 (although simultaneously recorded) revealed a reasonably good match regarding sleep fragmentation. Thus, we believe that this device has the ability to serve as a useful ambulatory device for detecting sleep fragmentation in patients’ homes, although home studies have not been performed in the present study. It should be stressed, however, that the criteria for detecting EEG arousals were rather arbitrarily determined by a task force of the American Sleep Disorders Association (ASDA) [5]. Much more data are required to study the best criteria of sleep fragmentation in predicting clinical outcomes. In fact, both the EEG and the autonomic arousals in our study were poorly (although statistically significantly) correlated with the ESS, probably indicating either poor subjective assessment by patients of their sleepiness level, or an objective limitation of the EEG arousal index and AAI prediction. This is not surprising, as previous studies have also failed to show good correlation between arousals from sleep and daytime sleepiness [15,26, 33]. As this study showed good agreement between the PAT defined AAI and the EEG defined arousals index, the poor correlation between the PAT-AAI and subjective daytime sleepiness assessment was also expected. When originally introduced, the ESS demonstrated good correlation with objective sleepiness measures (Multiple Sleep Latency Test, MSLT) [17], which later studies failed to replicate [34]. Also, it should be kept in mind that the sleep study, as well as MSLT, represents an acute state reflecting a specific point in time, while the ESS represents a trait measure correct for the recent period. Thus, the ability of the PAT to predict daytime sleepiness needs further evaluation, probably with more objective modalities such as MSLT. There are several limitations in this study. Although the purpose of the study was to evaluate an ambulatory device, the studies took place in the sleep lab because we wanted to assess at the outset the accuracy of the device in detecting arousals. Obviously this device will have to be studied in the home environment during a second stage. Second, we aimed at detecting sleep fragmentation but chose to compare autonomic arousals to the commonly detected EEG arousals. Per definition, we ignored EEG shifts that were shorter than 3 s, even if there was a clear prior flow limitation. It is not unlikely that we limit the PAT’s usefulness, and the potential augmentation of its clinical use, by comparing it only to AASM defined arousals when it may be sensitive to shorter arousals. Further studies of this methodology are needed to find an optimal marker of sleep fragmentation that can predict clinical outcome. Third, the population studied consisted primarily of patients with snoring/sleep apnea syndrome, and a few healthy volunteers. One could argue that in other populations (such as 211 insomniacs) arousals from sleep might be associated with different patterns of autonomic activation; expanding the population of the study will add more information regarding autonomic arousals. Despite these limitations, we believe that this study supports the concept that an ambulatory device measuring PAT at the finger can accurately assess sleep fragmentation. Autonomic arousals can be assessed in an objective, reproducible and time saving procedure using a computerized automated algorithm. References [1] Martin SE, Engleman HM, Deary IJ, Douglas NJ. The effect of sleep fragmentation on daytime function. Am J Respir Crit Care Med 1996; 153:1328–32. [2] Martin SE, Engleman HM, Kingshott RN, Douglas NJ. Microarousals in patients with sleep apnoea/hypopnoea syndrome. J Sleep Res 1997; 6:276–80. [3] Engleman HM, Cheshire KE, Deary IJ, Douglas NJ. Daytime sleepiness, cognitive performance and mood after continuous positive airway pressure for the sleep apnoea/hypopnoea syndrome. Thorax 1993;48:911–4. [4] Martin SE, Wraith PK, Deary IJ, Douglas NJ. The effect of nonvisible sleep fragmentation on daytime function. Am J Respir Crit Care Med 1997;155:1596–601. [5] American Sleep Disorders Association, EEG arousals: scoring rules and examples. A preliminary report from the Sleep Disorders Atlas Task Force of the American Sleep Disorders Association. Sleep 1992; 15:174–84. [6] Loredo JS, Clausen JL, Ancoli-Israel S, Dimsdale JE. Night-to-night arousal variability and interscorer reliability of arousal measurements. Sleep 1999;22:916– 20. [7] Drinnan MJ, Murray A, Griffiths CJ, Gibson GJ. Interobserver variability in recognizing arousal in respiratory sleep disorders. Am J Respir Crit Care Med 1998;158:358 –62. [8] Drinnan MJ, Murray A, White JE, et al. Automated recognition of EEG changes accompanying arousal in respiratory sleep disorders. Sleep 1996;19:296– 303. [9] Pillar G, Bar A, Shlitner A, et al. Autonomic arousal index (AAI): an automated detection based on peripheral arterial tonometry. Sleep 2003; in press. [10] White DP, Gibb TJ, Wall JM, Westbrook PR. Assessment of accuracy and analysis time of a novel device to monitor sleep and breathing in the home. Sleep 1995;18:115 –26. [11] Aubert-Tulkens G, Culee C, Harmant-Van Rijckevorsel K, Rodenstein DO. Ambulatory evaluation of sleep disturbance and therapeutic effects in sleep apnea syndrome by wrist activity monitoring. Am Rev Respir Dis 1987;136:851–6. [12] McCall WV, Erwin CW, Edinger JD, et al. Ambulatory polysomnography: technical aspects and normative values (review). J Clin Neurophysiol 1992;9:68– 77. [13] Stoohs R, Guilleminault C. MESAM 4: an ambulatory device for the detection of patients at risk for obstructive sleep apnea syndrome (OSAS). Chest 1992;101:1221–7. [14] Pitson D, Chhina N, Knijn S, et al. Changes in pulse transit time and pulse rate as markers of arousal from sleep in normal subjects. Clin Sci (Colch) 1994;87:269–73. [15] Pitson DJ, Stradling JR. Autonomic markers of arousal during sleep in patients undergoing investigation for obstructive sleep apnoea, their relationship to EEG arousals, respiratory events and subjective sleepiness. J Sleep Res 1998;7:53–9. [16] Argod J, Pepin JL, Smith RP, Levy P. Comparison of esophageal 212 [17] [18] [19] [20] [21] [22] [23] [24] [25] G. Pillar et al. / Sleep Medicine 4 (2003) 207–212 pressure with pulse transit time as a measure of respiratory effort for scoring obstructive nonapneic respiratory events. Am J Respir Crit Care Med 2000;162:87–93. Johns MW. A new method for measuring daytime sleepiness: the Epworth Sleepiness Scale. Sleep 1991;14:540–5. Rechtschaffen A, Kales A. A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Los Angeles, CA: Brain Information Service/Brain Research Institute, UCLA, 1968. AASM, Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in adults. Sleep 1999;22:667 –89. Bar A, Pillar G, Dvir I, et al. Validation of a portable device based on peripheral arterial tonometry (PAT) for unattended home sleep studies, submitted for publication. Mathur R, Douglas NJ. Frequency of EEG arousals from nocturnal sleep in normal subjects. Sleep 1995;18:330–3. Brooks III JO, Friedman L, Bliwise DL, Yesavage JA. Use of the wrist actigraph to study insomnia in older adults. Sleep 1993;16:151–5. Otsuka K, Otsuka K, Seto K, et al. Ambulatory ECG-respiration monitoring system (letter). Am Heart J 1983;106:1173–6. Tvinnereim M, Mateika S, Cole P, et al. Diagnosis of obstructive sleep apnea using a portable transducer catheter. Am J Respir Crit Care Med 1995;152:775 –9. Hirvonen K, Hasan J, Hakkinen V, et al. The detection of drowsiness and sleep onset periods from ambulatory recorded polygraphic data. Electroenceph clin Neurophysiol 1997;102:132 –7. [26] Stradling JR, Barbour C, Glennon J, et al. Prevalence of sleepiness and its relation to autonomic evidence of arousals and increased inspiratory effort in a community based population of men and women. J Sleep Res 2000;9:381 –8. [27] Davies RJ, Vardi-Visy K, Clarke M, Stradling JR. Identification of sleep disruption and sleep disordered breathing from the systolic blood pressure profile. Thorax 1993;48:1242–7. [28] Horner RL, Brooks D, Kozar LF, et al. Immediate effects of arousal from sleep on cardiac autonomic outflow in the absence of breathing in dogs. J Appl Physiol 1995;79:151 –62. [29] Somers VK, Dyken ME, Mark AL, Abboud FM. Sympathetic-nerve activity during sleep in normal subjects. N Engl J Med 1993;328: 303 –7. [30] Morgan BJ, Crabtree DC, Puleo DS, et al. Neurocirculatory consequences of abrupt change in sleep-state in humans. J Appl Physiol 1996;80:1627–36. [31] Hornyak M, Cejnar M, Elam M, Wallin BG. Muscle sympathetic nerve activity during sleep in man. Brain 1991;114:1281–95. [32] Okada H, Iwase S, Mano T, et al. Changes in muscle sympathetic nerve activity during sleep in humans. Neurology 1991;41:1961–6. [33] Kingshott RN, Engleman HM, Deary IJ, Douglas NJ. Does arousal frequency predict daytime function? Eur Respir J 1998;12:1264–70. [34] Chervin RD, Aldrich MS. The Epworth Sleepiness Scale may not reflect objective measures of sleepiness or sleep apnea. Neurology 1999;52:125–31.