Fundamental Research

Transcription

1 Sleep. 15(4): American Sleep Disorders Association and Sleep Research Society Fundamental Research Dynamics ofeeg Slow-Wave Activity and Core Body Temperature in Human Sleep After Exposure to Bright Light Christian Cajochen, *Derk-Jan Dijk and Alexander A. BorbeIy Institute of Pharmacology, University of Zurich, Zurich. Switzerland Summary: In seven subjects sleep was recorded after a single 3-hour (21- hours) exposure to either bright light (BL, approx. 2,5 lux) or dim light (DL, approx. 6 lux) in a crossover design. The latency to sleep onset was increased after BL. Whereas rectal temperature before sleep onset and during the first 4 hours of sleep was higher after BL than after DL, the time course of electroencephalographic (EEG) slow-wave activity (SWA, EEG power density in the range of Hz) in nonrapid eye movement sleep (NREMS) differed only slightly between the conditions. After BL, SW A tended to be lower than after DL in the first NREMS-REMS cycle and was higher in the fourth cycle at the time when the rectal temperature did not differ. The differences in SWA may have been due to a minor sleep-disturbing aftereffect of BL, which was followed by a rebound. The data are not in support of a close relationship between SWA and core body temperature. Key Words: Core body temperature-light-slowwave sleep-slow-wave activity-spectral analysis-sleep homeostasis. Until the early eighties it was assumed that light exposure does not affect the human circadian system (1). Lewy et al. (2) were the first to demonstrate that bright light (BL) (approx. 2,5 lux) suppresses the plasma melatonin level in humans. Subsequently, it was shown that the circadian rhythms of core body temperature, cortisol, urine output and alertness exhibit phase shifts after scheduled exposure to BL (3-5). For example, the rhythms of both body temperature and cortisol were phase delayed after repeated exposure to BL in the evening (4), whereas a repeated exposure to BL in the early morning resulted in an advance in the rise of body temperature (6). When the sleep-wake cycle was held constant, BL exposure in the morning advanced the circadian rhythms of melatonin and body temperature and shortened rapid eye movement sleep (REMS) latency, whereas BL exposure in the evening delayed these Accepted for publication March Address correspondence and reprint requests to A. A. BorbeIy, Institute of Pharmacology, University of Zurich, Gloriastr. 32, 86 Zurich, Switzerland. *Present address: Center for Circadian and Sleep Disorders Medicine, Harvard Medical School, Brigham and Women's Hospital, Boston, MA 2115, U.S.A. 337 rhythms (7). Taken together, the results indicate that light affects the phase of a single circadian pacemaker. Also, the findings could contribute to the understanding of light therapy for disorders in which a circadian pathophysiology is thought to be prevalent [e.g. certain sleep disorders (8), jet lag (9), shift work difficulties (1) and seasonal affective disorder (SAD) (11)]. The effects of BL exposure on sleep parameters and the sleep electroencephalogram (EEG) are not yet well documented. In most studies with repeated BL exposure in the morning or in the evening, only selected sleep parameters (e.g. sleep onset time, REMS latency and total sleep time) were analyzed (12,13). Light scheduled in the morning reduced sleep duration at the expense of REMS, whereas the EEG power density (in the range of Hz) was not affected (6,14). Effects of light on body temperature have been recently documented. Thus, Badia et al. (15) reported that a single exposure to BL in the evening caused an immediate elevation of tympanic temperature. We confirmed that a single 3-hour exposure to BL in the evening elevates core body temperature (16). This manipulation had an immediate effect that persisted for the first 4 hours of the subsequent sleep episode. Whereas the elevated rectal temperature was

2 338 C. CAJOCHEN ET AL. accompanied by an increase in sleep latency, no significant effect on the visually scored sleep stages was found. This finding was surprising as manipulations of body temperature both prior to sleep and during sleep have been shown to increase slow-wave sleep (SWS) (17-19). Also other authors have proposed a close nlationship between SWS and temperature regulation (2,21). To examine the repercussions oflight-induced temperature changes on sleep regulation, we analyzed the dynamics ofeeg slow-wave activity (SW A; EEG power density in the hz band), an indicator of sleep homeostasis, and its relation to core body temperature. Subjects and design METHODS The experiment was carried out in February and March at the Institute of Pharmacology, University tdf Zurich. Eight male subjects (age range years) were selected. All had regular sleep habits and were free of sleep complaints. The selection of subjects was based on a questionnaire in which the sleep habits, the medical history and the use of caffeine, alcohol and other drugs were assessed. Subjects with sleep complaints, a significant medical history or drug use were excluded from the study. The data of one subject could not be used for analysis for technical reasons. The experiment took place on three consecutive days. A first night in the sleep laboratory served for adaptation. On the evenings of the second and third day, subjects reported at 193 hours to the laboratory where the EEG, electromyogram (EMG) and electrooculogram (EOG) electrodes were attached, and a rectal probe was provided. From 21 to hours they were exposed to either BL (approx. 2,5 lux measured at eye level) or dim light (DL, approx. 6 lux) in a crossover design (three subjects were first exposed to BL; four first to DL). During the BL, subjects sat at a desk in front of a light screen (65 x 12 cm) consisting of eight white fluorescent tubes (Luxsana, true light with UV A and B, Duro-Test). The lower edge of the light screen was situated 8 cm above the floor. The distance between the subjects and the light source was not more than 1.5 m. They were allowed to read, play games or listen to music from a local radio station and were under continuous surveillance to prevent sleep. During the exposure period and the subsequent skep period, rectal temperature was recorded continuously. Every 3 minutes (between 2115 and 2345 hours) a 3-minute period was scheduled for recording the waking EEG. The subjects were requested to close their eyes for 1 minute and then to fixate a point on the wall for 2 minutes. After the 3-hour light exposure period the subjects went to bed at hour in a completely darkened and sound-attenuated room in the sleep laboratory. Polygraphic sleep recordings were obtained. Bed rest was terminated at 745 hours. EEG recording and analysis EEG, EOG and submental EMG were recorded on paper (paper speed of 1 mm/second) with a Grass 78-D polygraph. The amplitude frequency of the highpass filters for the EEG was set to.1 Hz, which corresponds to a time constant of.6 second. EEGs were derived from C3-A2 and C4-Al, but only one derivation (usually C3-A2) was used for the analysis. All signals were recorded with gold disc electrodes (Grass Instruments type E5GH), filled with electrode cream (EC2 Grass) and fixed with collodium. The data were stored on analog magnetic tape (Hewlett Packard 3968A). In addition, two EEGs, one EMG and one EOG were on-line digitized (sampling rate: 128 Hz) by a signal processor board (containing a TMS-32-1 chip of Texas Instruments) of a personal computer (PC; Olivetti-M24). Before digitizing, the EEG was low-pass filtered at 25 Hz (24 db/octave) and on-line processed by a Radix 8 FFT routine, which was implemented on the signal processor board. The spectra were computed with a rectangular window for 4-second epochs. To obtain a mean spectrum per 2 seconds, five 4-second spectra were averaged off-line. Four-second epochs, during which the limit of the AD converter was reached (usually due to artifacts caused by movements), were excluded. Further data reduction was achieved by collapsing values into.5-hz (in the range of Hz) or I-Hz ( Hz) bins. All EEG recordings were visually scored for 2-second epochs according to the criteria of Rechtschaffen and Kales (22). A time signal generated by the PC every 2 seconds allowed an accurate sychronization between power spectra and visual scores. The sleep scores were also stored in the pc. Rectal temperature recording Rectal temperature was continuously recorded during the exposure period and bed period. The temperature values were digitized and stored every 8 seconds by a portable temperature sampler (Minilog TA, V2). The temperature probe (diameter 5 mm, resolution.2 C) was inserted approximately 1 cm into the rectum. Statistics For statistical analyses (ANOV A, two-tailed t test) the SAS statistical package (SAS software, SAS Institute Inc., Cary, NC, U.S.A) was used.

3 LIGHT EFFECTS ON SLEEP EEG AND TEMPERATURE 339 TABLE 1. Sleep stages per NREMS-REMS cycle for the dim light (DL) and bright light (BL) condition a Cycle Condi- Total sleep Stage tion I episode W DL BL DL BL DL BL DL BL DL BL SWS DL BL REMS DL b BL b MT DL BL REMSL DL BL SL DL ' BL c TST DL BL SWA DL BL u Mean values (minutes) SEM (n = 7); W, waking; MT, movement time; SWS, stages 3 + 4; REMSL, REM sleep latency; SL, sleep latency; TST, total sleep time; SWA, slow-wave activity ( Hz) in NREMS (/LV2). h REMS episode 4 not completed in all subjects (see Methods).,. Significant differences between DL and BL (p <.5; paired t test). Sleep stages RESULTS Table 1 indicates the sleep stages for the four nonrapid eye movement sleep (NREMS)-REMS cycles and for the total sleep episode. The cycles were defined according to the criteria of Feinberg and Floyd (23), with two exceptions: for cycle 4, a minimum of 5-minute REMS following NREMS was required, and for cycle 2-4 the first 2-second interval after REMS was taken as the starting epoch of the cycle. The latency to sleep onset (i.e. the interval from lights off to the first occurrence of stage 2 or REMS) was significantly longer after the exposure to BL (BL: 29.7 ± 12.6; DL: 9.9 ± 1.5 minutes ± SEM; p <.5 paired t test on log-transformed values; the data were log transformed because the original data were not normally distributed), whereas the latency to REMS (i.e. the interval between sleep onset and the first occurrence of REM sleep) was slightly shorter but not significantly different from DL (BL: 49.4 ± 11.2; DL: 58.6 ± 12.1 minutes; p >.5). The midpoints of corresponding NREMS REMS cycles did not differ significantly between the conditions. Over the entire sleep episode, none of the sleep stages differed significantly between the conditions (p >.5; paired t test). The time course of SWS and REMS was analyzed with a two-factor ANOYA for repeated measures on both factors (cycle: 1-4; condition: DL, BL). For this analysis SWS and REMS were expressed as a percentage of total sleep time per NREMS-REMS cycle. The effect of cycle was significant for both SWS (F 3,I8 = 52.83; P <.1) and REMS (F 2 I8 = 5.31; P <.5). However, neither for SWS (F I 6 =.35) nor for REMS (F I,6 = 5.9) was a significant effect of condition observed (SWS: p >.2; REMS: p =.51). No significant interaction between the factors condition and cycle was present (SWS, F 3,I8 =.71; P >.5; REMS, F 2 I8 = 1.1; p >.1). Time course of EEG power density Because only four cycles were completed by all subjects, the analysis was limited to these cycles. The upper two panels of Fig. 1 illustrate the changes of EEG power density in NREMS for the BL and DL condition. As previously (24), the mean power density iii NREMS was expressed for each of the cycles 2-4 relative to

4 1-34 C. CAJOCHEN ET AL. >-,... en c:.! G> U C > G> NREMS-eplsodes Bright Light (Bll O-----'-l--lr---Ir----'I cycle 1. In both experimental conditions power density decreased from cycle 1 to 4, particularly in the delta and theta frequencies. The largest reduction was present in the Hz band. In the third panel of Fig. 1 the BL values are expressed relative to the corresponding DL values. In cycle 1 power density in the hz band was significantly lower in BL than in DL, whereas in cycle 4 the opposite changes were seen for the hz band. Significantly lower values in BL were observed also in the Hz band in cycle 3 and in the Hz and Hz bands in cycle 4. In REMS, significant reductions in BL occurred only in the first cycle ( Hz; Hz; Hz; data not shown). Over the entire sleep episode no significant changes were seen in the spectra ofnrems and REMS (bottom panel Fig. 1). -..I C c: -z:..2l..i c: CD... a. f!-.. c:.!! >- u ->c: u., C) " c:... " G> c: Co a. en G>... u o 5 Dim Light (DU NREMS-eplsodes (BUDl) ' Frequency (Hz) FIG. 1. Change of power density during NREMS over consecutive NREMS-REMS cycles after exposure to BL and DL (upper two panels). For each frequency bin, subject and condition spectral values are expressed'relative to the value in the first NREMS-REMS cycle (=1%). Geometric means are plotted. Lines above the abscissa ipdicate frequencies for which a one-way ANOV A on the log-transformed spectral values reveal a significant effect of cycle (p <.5). In the lower two panels spectral values of BL are expressed relative Dynamics of slow-wave activity and body temperature in NREMS-REMS cycles For a more detailed visualization of the time course ofswa and rectal temperature, each NREMS episode was subdivided into 2 5-percentiles, and each REMS episode into 4 25-percentiles regardless of the absolute episode duration (25). Mean values for SWA and rectal temperature were computed for each of these percentiles (Fig. 2). SW A exhibited a buildup in the first part ofnrems episodes and a rapid decline prior to the REMS episodes. Also, the typical declining trend of SW A over successive NREMS episodes was present (26). Although the first three REMS episodes in the BL condition showed an earlier onset than in the DL condition, the differences were not significant. Rectal temperature in DL was significantly lower than in BL during the last hour of the exposure time (Table 2) and in the first 4 hours of sleep (16). In the second part of sleep the differences became progressively smaller. For a statistical analysis of the change of SW A over cycles and its buildup within NREMS episodes, mean SWAin NREMS (SW Amean) and SW A averaged over the first 3 minutes of an NREMS episode (NREMS SW A3 min) were computed for each cycle. A repeatedmeasure ANOV A for each SW A parameter with factors time and condition revealed a significant effect of time and condition (p <.5, for both SW A parameters). Posthoc comparisons of BL and DL showed to the average value ofdl (=1%). Lines above the abscissa indicate frequency bins in which a significant difference between corresponding cycles (1-4) was present (paired t test on log-transformed data; p <.5).. Sleep, Vol. 15, No.4, -1992

5 LIGHT EFFECTS ON SLEEP EEG AND TEMPERATURE ). - u Q)... ::J -ro E (ij Q) a: rfi. 3 >-: 2 «- Q) > ro ;: 1 I :!= Ci5 o o :.! iii : : r o Hours FIG. 2. Dynamics of SWA and rectal temperature for BL (interrupted lines) and DL (continuous lines). Hatched areas indicate ± I SE. Mean percentiles (n = 7); SW A is expressed as the percentage of the mean NREMS value (1%). The bars above the abscissa indicate REMS episodes (REMS episode 4 is incomplete; open bars = BL; filled bars = DL). The curves connect mean values for percentiles and are plotted relative to the mean onset and termination 9f each cycle (see text). significant differences in cycle 4 of SW Amean and NREMS-SW A3 min (p <.5; paired t test). An analysis of consecutive 5-minute intervals revealed that the significant change in cycle 4 occurred in the last two 5-minute intervals of the first 3 minutes. A veraged over the entire sleep episode, SWAin NREMS did not differ significantly between conditions (DL = 1%; BL = 13.4%; Table O. The rise rate of SWAin the first 3 minutes in each of the four NREMS episodes (NREMS-SW Arise) was estimated by calculating for each individual the median of the differences in SW A over consecutive 5-minute intervals (24). A repeated measure ANOV A (two factors: condition, cycle) revealed a significant effect of cycle (F 3 18 = 21.17; P <.1), no effect of TABLE 2. Rectal temperature during dim light (DL) and bright light (BL) exposure prior to sleepa Time of day DL BL (.13) (.15) (.12) (.14) (.13) (.13) (.16) (.13) (.16) (.12)* (.15) 36.4 (.13)* a Mean values ("C) SEM per 3 minutes (n = 7). Significant differences between DL and BL indicated by *p <.1 (paired t test). condition (F1.6 =.41; p >.5) and no significant interaction (F3 18 =.36; P >.5). The mean rctal temperature per cycle (TEMPean) was higher in the first three cycles after BL than after DL (p <.1). The change of rectal temperature in the first 3 minutes of a NREMS episode was determined by computing the difference between the mean 5-minute value preceding the episode and the mean value of the 25-3-minute interval. There were no significant differences between BL and DL. EEG power density in the waking EEG The analysis of the waking EEG was limited to artifact-free epochs (59.5% of all epochs; artifacts were frequent due to movements and eye blinks). The EEG power density in the Hz range showed a trend toward lower values in the BL condition (eyes open) compared to DL (p <.1; data not shown). However, neither for the eyes open nor for the eyes closed condition were there any significant differences between BL and DL. DISCUSSION Exposure to BL in the evening increased core body temperature, a result that is in accordance with the findings of Badia et al. (15). This prominent effect on temperature was accompanied by only minor changes of sleep and the EEG. In contrast to the increase in sleep latency after BL, the REMS latency was not affected. Although the REMS latencies in DL were rather short, they are within the range of normal healthy young subjects when they are well adapted to. the laboratory (27). There were no significant differences in the sleep stages, and the typical decline of SW A over successive cycles was similar under both experimental conditions (Fig. 1). For the entire sleep episode, SW A as well as power density in the other frequency bands showed no significant differences between the conditions. Nevertheless, small but significant differences in the time course of SW A were present. In the first NREMS.,.. Sleep. Vol. 15. No

6 342 C. CAJOCHEN ET AL.. SWAme"o (%) TABLE 3. Slow-wave activity (SWA) and rectal temperature parameters/or the NREMS-REMS cycles Parameter NREMS-SWA3minb (%) NREMS-SWArie (%) Tempmean d ("C) Condition DL 189. (8.5) BL (16.5) DL (1.7) BL (17.2) DL 34.1(1.7) BL 36.5 (5.7) DL 35.8 (.1)** BL 36.1 (.1 )** DL -1.5 (7.2) BL (3.1) I Mean SWAin NREMS. b Mean SWAin the first 3 minutes of the NREMS episode. e The interindividual mean of the intraindividual median valul! of the differences in SW A over consecutive 5-minute intervals in the first 3 minutes of a NREMS episode. d Mean rectal temperature ("C) per cycle. " e Change in rectal temperature over the first 3 minutes of the NREMS episode (1O- 2 C); difference between the mean 5-minute value preceding the episode and the mean value of the 25-3-minute interval. All SWA parameters are expressed as a percentage of the mean SWA in NREMS (1%). Significant differences between BL and DL are indicated by *p <.5 and **p <.1 (paired t test). Cycle (12.6) 7.3 (9.1) 44.7 (4.1)* (14.9) 85.2 (5.) 64.6 (9.2)* 81.4 (11.5) 52.8 (7.1) 39. (3.1)* 94.8 (12.9) 64.2 (9.4) 51.7 (6.)* 2.8 (4.6) 1.9 (3.) 6. (.9) 19.9 (5.1) 11.6(1.7) 1.9 (2.4) 35.7 (.1)** 35.8 (.1)** 36. (.1) 36. (.1)** 36.1 (.1 )** 36. (.1).5 (1.9) 6.7 (2.7) 1.6 (3.4). (2.4) 2.4 (2.9) 1.5 (2.4) REMS cycle, power density in a frequency bin within the delta range was lower in BL than in DL. Conversely, an increase in delta activity was present in the fourth cycle. This shift in the distribution ofswa could have been caused by a slight suppression in the first cycle as a consequence of BL and a rebound during later parts of sleep. A similar intra sleep rebound has been induced after the selective suppression of SWAby acoustic stimuli in the first 3 hours of sleep (28). The higher rise rate of SW A in the fourth NREMS episode is an indication that "sleep pressure" in the later part of sleep was higher in BL than in DL (24). On the other hand, it is more difficult to specify the factor that led to the slight initial suppression of SWAin BL. Although the longer sleep latency is consistent with a sleep-disturbing aftereffect of BL, the rise rate in the first NREMS episode was not altered (Table 3). It may be the somewhat earlier initiation ofrems in BL (Fig. 2) that gave rise to a curtailment of SWAin the first cycle and a compensatory increase in cycle 4. The present study provided the opportunity of comparing the dynamics of SW A and rectal temperature. It had been reported that the elevation of body temperature during waking to values that are at the limit or above values that occur under physiological conditions enhance SWS (17-19). Other authors proposed that a regulated rapid decline in core body temperature after sleep onset is a necessary prerequisite for sustained SWS (29). The present data do not support the assumption of a close relationship between body temperature on the one hand and either SWS or SW A on the other hand. The higher body temperature in BL during the hours preceding sleep (16) did not affect SWS, and, if anything, reduced SW A in the first NREMS-REMS cycle. As has been mentioned, the slight increase of SWAin cycle 4 constituted probably a compensatory response. However, the possibility that it may have been due to a delayed effect of the heat load cannot be excluded. There was no indication that the rate of initial temperature decline was related to SWS or SW A. If anything, the slight reduction ofsw A in the first cycle of BL was associated with a steeper decline of rectal temperature (Table 3). This weak relationship between SWS/SW A and body temperature is in accordance with Dijk et al. (3) who found a dissociation between the time course ofsw A and body temperature. In conclusion, exposure to BL prior to sleep constitutes a subtle method for elevating body temperature at sleep onset and during the first part of sleep. This effect is mediated via the eyes rather than by the radiant heat of light (16) and must be due to an action on mechanisms subserving thermoregulation and/or circadian rhythmicity. Processes underlying sleep homeostasis are apparently little affected when body temperature is manipulated within a physiological range. Acknowledgements: We thank Dr. I. Tobler and Dr. P. Achermann for their comments and Dr. M. MUnch for her assistance in data acquisition. The study was supported by the Swiss National Science Foundation, grant no REFERENCES 1. Wever RA. Zur Zeitgeberstarke eines Licht-Dunkel-Wechsels flir die circadiane Periodik des Menschen. Fj/iigers Arch 197; 321:

7 '., LIGHT EFFECTS ON SLEEP EEG AND TEMPERATURE Lewy AJ, Wehr TA, Goodwin FK, Newsome DA, Markey SP. Light suppresses melatonin secretion in humans. Science 198; 21: Czeisler CA, Kronauer RE, Allan JS, Duffy JF, Jewett ME, Brown EN, Ronda JM. Bright light induction of strong (type ) resetting of the human circadian pacemaker. Science 1989;244: Czeisler CA, Allan JS, Strogatz SH, Ronda JM, Sanchez R, Rios CD, Freitag WO, Richardson GS, Kronauer RE. Bright light resets the human circadian pacemaker independent ofthe timing of the sleep-wake cycle. Science 1986;233: Wever RA, Polasek J, Wildgruber CM. Bright light affects human circadian rhythms. Pfliigers Arch 1983;396: Dijk DJ, Visscher CA, Bloem GM, Beersma DGM, Daan S. Reduction of human sleep duration after bright light exposure in the morning. Neurosci Lett 1987;73: Lewy AJ, Sack RL, Singer CM. Immediate and delayed effects of bright light on human melatonin production: shifting 'dawn' and 'dusk' shifts the dim light melatonin onset (DLMO). Ann NY A cad Sci 1985;453: Lewy AJ, Sack RL, Miller S, Hoban TM. Antidepressant and circadian phase-shifting effects oflight. Science 1987;235: Daan S, Lewy AJ. Scheduled exposure to daylight: a potential strategy to reduce "Jet Lag" following transmeridian flight. Psychopharmacol Bull 1984;2: Czeisler CA, Johnson MP, DuffY JF, Brown EN, Ronda JM. Exposure to bright light and darkness to treat physiologic maladaptation to night work. N Engl J Med 199;322: II. Rosenthal NE, Levendosky AA, Skwerer RG, Vanderpool JRJ, Kelly KA, Hardin T, Kasper S, DellaBella P, Wehr TA. Effects oflight treatment on core body temperature in seasonal affective disorder. BioI Psychiatry 199;27: Sack RL, Lewy AJ, Miller S, Singer CM. Effects of morning versus evening bright light exposure on REM latency. BioI Psychiatry 1986;21: Drennan M, Kripke DF, Gillin Je. Bright light can delay human temperature rhythm independent of sleep. Am J Physiol 1989; 257:RI Dijk DJ, Beersma DGM, Daan S. Bright morning light advances the human circadian system without affecting NREM sleep homeostasis. Am J PhysioI1989;256:RI Badia P, Culpepper J, Boecker M, Myers B, Harsh J. The immediate effect of bright and dim light on body temperature. Sleep Res 199; 19: Dijk DJ, Cajochen C, Borbely AA. Effect of a single 3-hour exposure to bright light on core body temperature and sleep in humans. Neurosci Lett 1991;121: Bunnel DE, Agnew JA, Horvath SM, Jopson L, WillsM. Passive body heating and sleep: influence of proximity to sleep. Sleep 1988:11: Home JA, Moore VJ. Sleep EEG effects of exercise with and without additional body cooling. Electroencephalogr Clin Neurophysio/ 1985;6: Home JA, Reid AJ. Night-time sleep EEG changes following body heating in a warm bath. Electroencephalogr Clin Neurophysiol 1985;6: Berger RJ, Palca JW, Walker JM, Phillips NH. Correlation between body temperature, metabolic rate and slow wave sleep in humans. Neurosci Lett 1988;86: Szymusiak R, McGinty D. Control of slow-wave sleep by thermoregulatory mechanisms. In: Issa FG, Suratt PM, Remmers JE, eds. Sleep and respiration. New York: Wiley-Liss, 199: Rechtschaffen A, Kales A. A manual of standardized terminology techniques and scoring systems for sleep stages of human subjects. US Department of Health, Education and Welfare, Public Health Service, Bethesda, MD, Feinberg I, Floyd Te. Systematic trends across the night in human sleep cycles. Psychophysiology 1979; 16: Dijk DJ, Brunner DP, Borbely AA. Time course ofeeg power density during long sleep in humans. Am J Physiol 199;258: R Brunner DP, Dijk DJ, Borbely AA. A quantitative analysis of phasic and tonic submental EMG activity in human sleep. Physiol Behav 199;48: Achermann P, Borbely AA. Simulation of human sleep: ultradian dynamics of electroencephalographic slow-wave activity. J BioI Rhythms 199;5: Belyavin A, Nicholson AN. Rapid eye movement sleep in man: modulation by benzodiazepines. Neuropharmacology 1987;26: Dijk DJ, Beersma DGM, Daan S, Bloem GM, Van den Hofdakker RH. Quantitative analysis of the effects of slow wave sleep deprivation during the first 3 h of sleep on subsequent EEG power density. Eur Arch Psychiatry Neural Sci 1987;236: Sewitch DE. Slow-wave sleep deficiency insomnia: a problem in thermo-downregulation at sleep onset. Psychophysiology 1987; 24: Dijk DJ, Cajochen C, Tobler I, Borbely AA. Sleep extension in humans: sleep stages, EEG power spectra and body temperature. Sleep 1991;14: