Sleep Research Online 2(3): 65-69, 1999http://www.sro.org/1999/Cajochen/65/ Printed in the USA. All rights reserved.

Correspondence:Christian Cajochen, Ph.D., Brigham & Women’s Hospital, Circadian, Neuroendocrine and Sleep Disorders Section,Endocrine-Hypertension Division, Harvard Medical School, 221 Longwood Avenue, Boston, MA 02115, USA, Tel: 617-732-7991, Fax:617-732-4015, E-mail: ccajochen@gcrc.bwh.harvard.edu.

1096-214X© 1999 WebSciences

During the transition from wakefulness or REM sleep tononREM sleep the electroencephalogram changes from a lowamplitude high frequency pattern to a high amplitude lowfrequency pattern. This high amplitude low frequency patternexhibits distinct dynamics across the nonREM-REM sleepcycle and across the nocturnal sleep episode in humans(Aeschbach and Borbély, 1993; Dijk et al., 1990). Recentlysome of the neurophysiologic processes underlying thesechanges have been identified (for a review see Amzica andSteriade, 1998; McCormick and Bal, 1997).

Computerized analyses of the EEG during sleep havedemonstrated that the power in delta and theta frequencies(0.75-7.0 Hz) components and in particular slow-wave activity(SWA; EEG power density in the 0.75-4.5 Hz band) declinesacross the sleep episode and increases with the duration ofwakefulness preceding sleep (Borbély et al., 1981; Dijk et al.,1987) almost independent of circadian phase (Dijk et al.,1997). In the classical visual staging of human nonREM sleepthese phenomena are reflected in the decline of slow wavesleep in the course of sleep and the increase of slow wave sleepafter sleep deprivation (Webb and Agnew, 1971).

Most of the aforementioned sleep studies in humans werebased on a single central EEG derivation. It is well known thatthe amplitude of slow waves is highly dependent upon theposition of the EEG derivation. However, it is not knownwhether activation of sleep regulatory mechanisms by anextension of wakefulness results in regional differences of therelative increase of low frequency components in the EEG(0.75-7.0 Hz). Earlier topographic studies of the dynamics ofSWA during the transitions from stage 1, 2, 3 to 4 or theta andalpha activity during the transition from wakefulness to sleepreported local differences (Buchsbaum et al., 1982; Wright JR

et al., 1995). Furthermore, distinct EEG frequency bins havebeen shown to exhibit prominent shifts in power along theantero-posterior axis during baseline sleep (Werth et al., 1996),and interhemispheric asymmetries in the nonREM sleep EEGwere observed after unilateral activation of the motor cortexduring wakefulness (Kattler et al., 1994). Finally, PET studieshave demonstrated that the decline of cerebral blood flow(rCBF) during SWS is most prominent in frontal cortical areas(Maquet et al., 1997; Hofle et al., 1997; Braun et al., 1997). Allthese data may indicate topographical differences in thedynamics of the EEG and sleep regulatory processes. Tofurther investigate topographical aspects of the EEG duringnonREM sleep subjects underwent sleep deprivation, whichhas previously been shown to increase delta and theta activityrecorded from central derivations (Borbély et al., 1981; Dijk etal., 1993). The dynamics of the human EEG along the antero-posterior axis during both baseline sleep and recovery sleepwere quantified. We hypothesized that the increase in lowfrequency components of the EEG as induced by sleepdeprivation would be most prominent in frontalderivations.

METHODS

SubjectsFour men and two women (mean age: 21.8 years; range: 19-

28 years) were studied. Women were studied in the follicularphase of the menstrual cycle. All subjects were free frommedical, psychiatric, and sleep disorders as assessed byhistory, a physical examination and questionnaires. Subjectswere instructed to abstain from caffeine, nicotine, alcohol, anddrugs for the 3 weeks before their study; their compliance was

The effect of sleep deprivation (40 h) on topographic and temporal aspects of electroencephalographic (EEG) activityduring sleep was investigated by all night spectral analysis in six young volunteers. The sleep-deprivation-induced increaseof EEG power density in the delta and theta frequencies (1-7 Hz) during nonREM sleep, assessed along the antero-posterioraxis (midline: Fz, Cz, Pz, Oz), was significantly larger in the more frontal derivations (Fz, Cz) than in the more parietalderivations (Pz, Oz). This frequency-specific frontal predominance was already present in the first 30 min of recovery sleep,and dissipated in the course of the 8-h sleep episode. The data demonstrate that the enhancement of slow wave EEG activityduring sleep following extended wakefulness is most pronounced in frontal cortical areas.

CURRENT CLAIM:

Frontal Predominance of a Relative Increase in Sleep Deltaand Theta EEG Activity after Sleep Loss in Humans

Christian Cajochen, Rebecca Foy and Derk-Jan DijkCircadian, Neuroendocrine and Sleep Disorders Section, Division of Endocrinology, Brigham and Women's

Hospital, Harvard Medical School, Boston, Massachusetts 02115, USA

The activation of sleep regulatory processes by an extension of wakefulness from 16 to 40 h results ina relative increase of low frequency EEG activity that is most pronounced in frontal cortical areas.

verified on the day of admission to the laboratory withurinary toxicologic analysis. Subjects were instructed to keepa regular sleep-wake schedule (bedtimes and waketimeswithin ±30 minutes of self-selected target times) during thethree weeks prior to their admission to the laboratory.Adherence to a regular schedule during the weekimmediately prior to admission was verified with a wristactigraph (Mini Motionlogger, AMI, Ardsley, NY, USA). Allsubjects gave their written informed consent. The protocol,consent form and advertisements were approved by theHuman Research Committee of the Brigham and Women'sHospital.

ProtocolFollowing three scheduled days and nights in the laboratory

during which subjects slept at their habitual bedtimes, thesubjects underwent a 40-h sleep deprivation under constantroutine conditions (Czeisler et al., 1985). Polysomnographicrecordings were taken during each 8-h sleep episode whichwere timed in such a way that they were centered at themidpoint of the subject's habitual sleep episode as assessed bywrist actigraphy and sleep logs taken one week prior to thestudy. The present analysis is based on the third baseline sleepepisode and the first sleep episode following sleep deprivation.

Sleep EEG recording and analysisThe EEG was recorded from Fz, Cz, Pz and Oz,

referenced against linked mastoids (A1, A2), together withthe two electrooculograms (EOGs), one ECG and oneEMG. All signals were on-line digitized (12 bit ADconverter, 0.122 µV/bit; storage sampling rate at 128 Hzfor the EEG) and digitally low-pass filtered at 60 Hz (4thorder Bessel type anti-aliasing filters, 24 dB/Oct.) Theywere then high-pass filtered using a time constant of 1.0-s(Vitaport-2 digital recorder, [TEMEC Instruments B.V.,Kerkrade, The Netherlands]). The raw signals were storedon-line on a Flash RAM Card (SanDisk, Sunnyvale, CA,USA) and downloaded off-line to an Apple (PowerMacintosh 7300/180) hard drive. The EEG signals duringthe 8-h sleep episodes were visually scored per 20-s epochaccording to Rechtschaffen and Kales (1968) with the

exception that the Cz-derivation was used for scoring(Vitaport Paperless Sleep Scoring Software [TEMECInstruments B.V., Kerkrade, The Netherlands]). Incomparison to a C3/C4 derivation, scoring sleep from theCz-derivation was very similar. Four-second epochs wereoff-line subjected to spectral analysis using a fast Fouriertransform (FFT, 10% cosine window) resulting in a 0.25Hz bin resolution. Before calculating the FFT, the EEGsignals were pre-whitened. All EEGs were manuallyscanned for artifacts due to body movements, sweating,eye movements, electrode malfunction, and stored in a file(1-s resolution). Artifact-free 4-s epochs were averaged over20-s epochs in order to align spectral data with sleep staging.

StatisticsThe statistical package SAS® (SAS® Institute Inc., Cary,

NC, Version v6.12) was used. Two-way and three-wayanalyses of variance for repeated measures (rANOVA) with thefactors 'night' (baseline, recovery), 'location' (Fz, Cz, Pz, Oz)and/or 'time interval' (Quarter of Sleep Episode, or Two MinuteInterval) were performed for each power value in eachfrequency bin separately. All p values derived from rANOVAswere based on Huynh-Feldt's (H-F) corrected degrees offreedom, but the original degrees of freedom are reported.Post-hoccomparisons using Duncan's multiple range- or two-sided paired t -tests were performed.

RESULTS

Sleep stagesTable 1 summarizes sleep parameters in the baseline night

and recovery night. Total sleep time, sleep efficiency (SE=total sleep time/time in bed*100) were significantly increasedafter extended wakefulness (Total sleep time: F1,5=6.6, p<0.05;Sleep efficiency: F1,5=6.6, p<0.05; rANOVA). Slow wavesleep (stage 3 and 4) was increased and a combined arousalmeasure (movement time+waking+stage1) as well as REMsleep were reduced (arousal: F1,5=26.7, p< 0.004; REM sleep:F1,5=7.9, p<0.05). Sleep latency appeared to be reduced(F1,5=3.9, p<0.1), whereas REMS latency was not affected(p>0.4).

66 CAJOCHEN ET AL.

Table 1Sleep Parameters for the Baseline Night and the Recovery Night.

Sleep Parameter Baseline night Recovery nightTotal sleep time (min) 438.6 ± 6.0 458.5 ± 5.0*

Sleep efficiency (%) 91.4 ± 1.2 95.5 ± 1.1*

Sleep latency (min) 16.9 ± 4.6 9.5 ± 2.3#

REM sleep latency (min) 58.0 ± 2.1 60.7 ± 7.7

Stage 2% 47.7 ± 1.5 48.2 ± 1.2

REM sleep% 24.5 ± 1.8 19.6 ± 2.7*

Slow wave sleep% 17.8 ± 1.1 25.5 ± 1.9**

Stage1+Waking+Movement Time% 20.7 ± 2.6 9.9 ± 1.4**

Mean values with SEM (n=6) are expressed as percentage of total sleep time. #p<0.1; *p<0.05; **p<0.004

EEG power densityAll-night EEG power spectra during nonREM sleep were

calculated for each derivation (Fz, Cz, Pz, Oz) during thebaseline and recovery sleep episode (Fig. 1 top panel). A two-way rANOVA with the factors night and location revealed asignificant main effect for the factor night in the frequencyranges from 1-7 Hz and 13.5-14 Hz (p<0.05 in all cases),reflecting the increase in the delta and theta activity and thereduction in spindle activity. In addition, a main effect for thefactor location was present in the frequency ranges 0.75-6.25Hz, 10.5-12.5 Hz, 13.25-14.75 Hz and 16.25-20 Hz (p<0.05).Significant interactions between the factors night and locationwere found for frequency bins ranging from 0.75-5.5 Hz andfrom 10.25-10.5 Hz (p<0.05). In order to visualize therepercussions of 40 h of sleep deprivation on the full spectrum,EEG during nonREM sleep, for each frequency bin, EEGpower density obtained during the recovery sleep episode wasexpressed as a percentage of the corresponding value from thebaseline sleep episode (Fig. 1 bottom panel). A one-wayrANOVA with the factor location was significant for thefrequency bins between 1-1.75 Hz, 3-3.25 Hz and the 9.5-9.75

Hz-bin (p<0.05). Post hoccomparisons of relative EEGpower density in the recovery vs. baseline night for eachderivation revealed significant higher EEG power densityafter sleep deprivation in the range of 0.75-9 Hz and 9.75-11.5 Hz and significant lower values in the 13.25-13.5 Hzbin for the frontal derivation. For the central derivationhigher values in the range of 0.75-8.5 Hz and 10-10.5 Hzand lower values in the 13.5-13.75 Hz bin were observed. Inthe parietal derivation, significant higher values were foundin the 2-8.5 Hz range and in the occipital derivation in the 5-7 Hz range.

The temporal evolution of SWA during nonREM sleep(stages 2, 3, 4) was computed for each quarter (2 h-interval) ofthe baseline and recovery sleep episode (Fig. 2). An increase inSWA after sleep deprivation was observed for all locations. Athree way rANOVA with factors night, location and timerevealed a significant main effect for the factors night (F1,5=30.5; p<0.003), location (F3,15=22.6; p<0.002) and time (F3,15=26.3; p<0.002) and a three-way interaction between the factorsnight x location x time (F9,45=8.55; p<0.006). SWA in thefrontal derivation was significantly higher in the first, the

67FRONTAL DOMINANCE OF SLEEP DEPRIVATION EFFECTS

Figure 1. Top panel: Absolute power spectra (log scale) for nonREM sleep (stages 2, 3, and 4) in the frontal (Fz), central (Cz), parietal (Pz) andoccipital derivation (Oz) during baseline and recovery sleep after total sleep deprivation (40 h). Horizontal symbols and lines at the bottom ofthe first panel indicate frequency bins for which the factors: night, location and the interaction night x location was significant (two-way ANOVAfor repeated measures, p<0.05). Bottom panel: Effects of sleep deprivation on the power spectrum in non-REM sleep (stages 2, 3, and 4). Datawere averaged over the first 7.5 h of recovery sleep and were expressed as percentage (+1 SEM) of the corresponding value in baseline sleep.Horizontal symbols indicate frequency bins for which the value in recovery sleep was significantly different from the baseline value (paired t-test on log transformed values p<0.05).

second and third quarter of the sleep episode. In the centralderivation, SWA was higher in the first and second quarter,whereas for the parietal derivation significant higher SWAvalues were only found in the first quarter (Duncan's multiplerange test p<0.05).

The effect of sleep deprivation on the evolution of SWA wasanalyzed for the first 36 min of the first nonREM episode witha time resolution of 2 min (Fig. 3). A three-way rANOVA withfactors night, location and time (30 two-min intervals)demonstrated a significant effect for the factors night (F1,15=13.2; p<0.02), location (F3,15=13.5; p<0.02), and time (F14,70=21.9; p<0.001) as well as a significant three-way interactionbetween the factors night x location x time (F42,240=2.4; p<0.05). The results of this ANOVA reflect a more pronounced

rise of SWA in the more frontal derivations as compared to theparietal derivations in the very beginning of recovery sleep.

DISCUSSION

The present analyses confirm our hypothesis that theincrease in low frequency components of the EEG as inducedby sleep deprivation is more pronounced in frontal regions ofthe brain. In particular, two indexes of sleep pressure, theglobal decline of slow-wave activity in the course of anocturnal sleep episode after sleep deprivation and thedynamics of the initial rise of SWA after sleep deprivation wereboth predominant in the frontal derivation. Furthermore, theenhancement in low frequency activity after sleep deprivation

68 CAJOCHEN ET AL.

Figure 3. Evolution of slow-wave activity during the first 36 min after sleep onset (first occurrence of stage 2) in baseline sleep and recoverysleep derived from a frontal (Fz), central (Cz), parietal (Pz) and occipital (Oz) derivation. Data were averaged per 2-min intervals (n=6; +1 SEM).The first data point in each panel represents EEG slow-wave activity during wakefulness and stage 1 before sleep onset. Dashed vertical linesdelineate sleep onset. For statistics see text.

Figure 2. Dynamics of slow-wave activity (SWA, EEG power density in the 0.75-4.5 Hz band) during nonREM sleep across quarters (2h-intervals) of the baseline and recovery sleep episode in the frontal (Fz), central (Cz), parietal (Pz) and occipital (Oz) derivation (n=6; +1 SEM).Asterisks indicate significant differences (p<0.05, baseline vs. recovery, Duncan's multiple range test).

comprised a broader frequency range and was significantlylarger in frontal derivation than in the parietal and occipitalderivations.

The observed increases in delta and theta activity and thedecrease in spindle activity are in good accordance withprevious studies in which the effect of sleep deprivation onEEG power spectra were quantified (Borbély et al., 1981; Dijket al., 1993) However, our current data for the first timedemonstrate that the total sleep-deprivation-induced increasein delta and theta varies along the antero-posterior axis. Inaddition, we have recently obtained data to indicate that theeffects of sleep loss on low frequency components in the EEGduring extended wakefulness may be more prominent in frontalareas than in posterior areas (Cajochen et al., 1998). These dataare in accordance with the concept that frontal cortical areas inparticular are affected by sleep deprivation (for a review seeHorne, 1993). Recent support for this concept may also bederived from the observation that rCBF in the anterior cingulateand orbitofrontal cortex as assessed by PET are negativelycorrelated with EEG delta activity during sleep (Hofle et al.,1997). A detailed analysis of the EEG derived from multiplebipolar derivations is needed for a more precise localization ofthe predominant increase in SWA after sleep deprivation.

ConclusionThese data indicate that some aspects of human sleep

regulation are local in nature as was previously demonstratedin dolphins (Oleksenko et al., 1992) and monkeys (Pigarev etal., 1997) and may imply that the human sleep EEG exhibts usedependent characteristics as hypothesized by Krueger and Obál(1993).

ACKNOWLEDGMENTS

We thank the subject volunteers and research technicians fortheir help. We also thank Ms. J. Jackson, Mr. E. Riel and Mr.J.M. Ronda for technical support. In addition, we thank Drs. K.P. Wright Jr. and Sat Bir S. Khalsa for comments on themanuscript. This research was supported by the NASACooperative Agreement NCC 9-58 with the National SpaceBiomedical Research Institute. C. Cajochen was in partsupported by a Swiss National Foundation Grant #823A-046640. Experiments were conducted in a General ClinicalResearch Center (grant M01 RR02635).

REFERENCES

1 Aeschbach D, Borbély AA. All-night dynamics of the humansleep EEG. Journal of Sleep Research 1993; 2: 70-81.

2. Amzica F, Steriade M. Electrophysiological correlates ofsleep delta waves. Electroencephalography and clinicalNeurophysiology 1998; 107: 69-83.

3. Borbély AA, Baumann F, Brandeis D, Strauch I, Lehmann D.Sleep deprivation: effect of sleep stages and EEG powerdensity in man. Electroencephalography and clinicalNeurophysiology 1981; 51: 483-93.

4. Braun AR, Balkin TJ, Wesenten NJ, Carson RE, Varga M,Baldwin P, Selbie S, Belenky G, Herscovitch P. Regional

cerebral blood flow throughout the sleep-wake cycle. AnH2(15)O PET study. Brain 1997; 120: 1173-7.

5. Buchsbaum MS, Mendelson WB, Duncan WC, Coppola R,Kelsoe J, Gillin JC. Topographic cortical mapping of EEGsleep stages during daytime naps in normal subjects. Sleep1982; 5: 248-55.

6. Cajochen C, Khalsa SBS, Wyatt JK, Czeisler CA, Dijk D-J.EEG and ocular correlates of circadian melatonin phase andhuman performance decrements during sleep loss.American Journal of Physiology 1999; 277, (in press).

7. Czeisler CA, Brown EN, Ronda JM, Kronauer RE,Richardson GS, Freitag WO. A clinical method to assess theendogenous circadian phase (ECP) of the deep circadianoscillator in man. Sleep Research 1985; 14: 295.

8. Dijk D-J, Beersma DGM, Daan S. EEG power density duringnap sleep: reflection of an hourglass measuring the durationof prior wakefulness. Journal of Biological Rhythms 1987;2: 207-19.

9. Dijk D-J, Brunner DP, Borbély AA. Time course of EEGpower density during long sleep in humans. AmericanJournal of Physiology 1990; 258: R650-61.

10. Dijk D-J, Hayes B, Czeisler CA. Dynamics of electroen-cephalographic sleep spindles and slow wave activity in men:effect of sleep deprivation. Brain Research 1993; 626: 190-9.

11. Dijk D-J, Shanahan TL, Duffy JF, Ronda JM, Czeisler CA.Variation of electroencephalographic activity during non-REM and REM sleep with phase of circadian melatoninrhythm in humans. J Physiol (London) 1997; 15: 505: 851-8.

12. Hofle N, Paus T, Reutens D, Fiset P, Gotman J, Evans AC,Jones BE. Regional cerebral blood flow changes as afunction of delta and spindle activity during slow wave sleepin humans. Journal of Neuroscience 1997; 17: 4800-8.

13. Horne JA. Human sleep, sleep loss and behaviour:Implications for the prefrontal cortex and psychiatricdisorder. British Journal of Psychiatry 1993; 162: 413-9.

14. Kattler H, Dijk D-J, Borbély AA. Effect of unilateralsomatosensory stimulation prior to sleep on the sleepEEG in humans. Journal of Sleep Research 1994; 3: 159-64.

15. Krueger JM, Obál F. A neuronal group theory of sleepfunction. Journal of Sleep Research 1993; 2: 63-9.

16. Maquet P, Degueldre C, Delfiore G, Aerts J, Pèters J-M,Luxen A, Franck G. Functional neuroanatomy of humanslow wave sleep. Journal of Neuroscience 1997; 17: 2807-12.

17. McCormick DA, Bal T. Sleep and arousal: thalamocorticalmechanisms. Annual Review of Neuroscience 1997; 20: 185-215.

18. Oleksenko AI, Mukhametov LM, Polyakova IG, Supin AY,Kovalzon VM. Unihemispheric sleep deprivation inbottlenose dolphins. Journal of Sleep Research 1992; 1: 40-4.

19. Pigarev IN, Nothdurft HC, Kastner S. Evidence forasynchronous development of sleep in cortical areas.NeuroReport 1997; 8: 2557-60.

20. Rechtschaffen A, Kales A, eds. A Manual of StandardizedTerminology, Techniques and Scoring System for SleepStages of Human Subjects. Bethesda MD: US Department ofHealth, Education and Welfare, Public Health Service, 1968.

21. Webb WB, Agnew HW. Stage 4 sleep: influence of timecourse variables. Science 1971; 174: 1354-6.

22. Werth E, Achermann P, Borbély AA. Brain topography of thehuman sleep EEG: Antero-posterior shifts of spectralpower. Neuroreport 1996; 20: 123-7.

23. Wright KP JR., Badia P, Wauquier A. Topographical andtemporal patterns of brain activity during the transition fromwakefulness to sleep. Sleep 1995; 18: 880-9.

69FRONTAL DOMINANCE OF SLEEP DEPRIVATION EFFECTS