Circadian Rhythms and Sleep Phase Advance in Moderately Sleepy ...

Sleep, 19(5):417-422 © 1996 American Sleep Disorders Association and Sleep Research Society

Circadian Rhythms and Sleep

Phase Advance in Moderately Sleepy and Alert Normals Timothy Roehrs, Raphael Salin-Pascual, Lori Merlotti, Leon Rosenthal and Thomas Roth Henry Ford Hospital, Sleep Disorders and Research Center, Detroit, Michigan, U.S.A.

Summary: Phase advanced sleep (by 4 hours) was studied in 28 healthy, normal men and women, aged 21-50 years, without nocturnal sleep or daytime sleepiness complaints. Eleven subjects (6 men and 5 women) with moderately short ($10 minutes) average daily sleep latencies on the multiple sleep latency test (MSLT) were compared to 17 (11 men and 6 women) with relatively long (~12 minutes) latencies. Nocturnal sleep on both a baseline and a shift night differed between the groups. The moderately "sleepy" group had shorter sleep latencies and less wakefulness during sleep than the moderately "alert" group. The phase advanced sleep schedule reduced sleep efficiency in both groups, but the moderately sleepy group showed a lessened effect of the shift. Additionally, sleep efficiency was reduced in the moderately sleepy group only during the first 2 hours of the sleep period, while sleep efficiency was uniformly disturbed in the moderately alert group throughout the night during the phase advance. Key Words: MSLT-Healthy normals-Phase advance-Nocturnal sleep efficiency.

Some healthy, normal young adults, reporting 7-8 hours of sleep each night with no nocturnal sleep difficulties and no complaints of daytime sleepiness, have unusually short sleep latencies on the multiple sleep latency test (MSLT) (1). Their average daily sleep latencies are :=;6 minutes, which is a level of sleepiness consistently seen in patients with disorders of excessive daytime sleepiness (2). It has been hypothesized that such individuals have accumulated a sleep debt resulting from chronically insufficient sleep relative to their biological sleep need (3). Given the hypothesized accumulated sleep debt such individuals have been termed "sleepy" normals. A number of findings converge to validate the description of such individuals as being sleepy. First, data from within-subject designs indicate that extended bedtime in sleepy normals is associated with a lengthening of average daily sleep latency (4-6). Normals with a reported 7.2 hours habitual bedtime and an initial average daily sleep latency of 3.8 ::!:: 1.2 minutes showed an increase in latency to 10 ::!:: 3.2 minutes after six consecutive nights of 10 hours bedAccepted for publication March 1996. Address correspondence and reprint requests to: Timothy Roehrs, Ph.D., Henry Ford Hospital, 2921 West Grand Blvd., Detroit, MI 48202, U.S.A.

time (5). In another study, normals reporting 7.5 hours nightly sleep on 2-week sleep logs and with an initial average daily sleep latency of 4.8 ::!:: 0.9 minutes had an average daily sleep latency of 12.3 ::!:: 5.2 minutes after 14 nights of 10 hours bedtime, while similarly sleepy normals maintained on their habitual 7.S hours sleep schedule did not improve. Their average daily sleep latency remained at 4.8 ± 3.0 minutes after the 14 nights on their habitual schedule (6). Secondly, data from between-subject designs show that the daytime sleepiness of sleepy normals is associated with predictable differences in nocturnal sleep efficiency, arousal threshold, and recovery sleep relative to "alert" normals. Sleepy normals had nocturnal sleep efficiencies of 95% compared to the 87% efficiencies of alert normals, defined as persons with average daily sleep latencies on the MSLT of 2::16 minutes (7). The sleepy subjects fell asleep more quickly at night, had less wake time during their sleep period, and when arousing they entered stage 1 sleep, rather than awakening compared to the alert normals. Also the sleepy subjects showed poorer performance on daytime psychometric testing. Next, sleepy subjects when forced to remain in bed continuously for 32 hours over a period including two nocturnal sleep periods (i.e. 2300-0700 hours) slept for 21.4 hours,

417

T. ROEHRS ET AL.

418

TABLE 1.

Inclusion criteria n per group Sex (mlO Mean age (years) Mean MSLT (minutes) Reported bedtime (p.m.) Reported sleep time (hours)

Study groups

Moderately alert

Moderately sleepy

MSLT 2:12 minutes 17 1116 30.6 (9.1) 16.8 (3.3) 10:57 (6 minutes) 7.7 (0.9)

MSLT $10 minutes 11 6/5 29.2 (10.9) 7.1 (2.1) 10:51 (14 minutes) 7.5 (0.8)

Data are means (:!:SD).

while alert normals under a similar enforced bedtime slept only 17.5 hours (8). Finally, arousal thresholds during sleep in sleepy subjects were higher than those of alert subjects and no different than those of a similar group of alert subjects who were randomly chosen to undergo 24 hours sleep deprivation and become sleepy (9).

Given the hypothesized accumulated sleep debt of individuals termed sleepy, their sleep efficiency should be somewhat resistant to challenges. Any number of studies have shown that shifting sleep schedules in normals is associated with transient sleep disturbances. A 180 degree sleep phase shift (bedtime shift from 2400 to 1200 hours) was associated with increased wakefulness during the last 4 hours of the 8-hour sleep period (10). Less severe sleep schedule shifts (i.e. 36 hours) also will produce increased wakefulness during the sleep period (11). The direction of the shift seems important as well; phase advancing, as opposed to phase delaying bedtime, produces greater sleep dis20 18

16 c

I

(j)

Cii

Twenty-eight healthy young men and women, aged 21-35 years (11 moderately sleepy and 17 moderately alert subjects as defined below, see Table 1) served in this study. All subjects had self-described normal sleep, no complaints of daytime sleepiness, and no habitual napping. They were healthy and drug free based on the screening described below. The protocol was approved by the institutional human rights committee and each subject signed an informed consent and was paid for his/her participation. Procedure

___ Mod Alert

12

Q. (j) (j)

Subjects

- . - Mod Sleepy

-1

(Jj

METHODS

14

>()

c

ruption (12). Thus, shifting sleep schedules in otherwise healthy normals is associated with transient reductions in sleep efficiency. The purpose of the present study was to extend the findings of nocturnal sleep differences in sleepy vs. alert normals to individuals who have moderate levels of sleepiness and alertness as defined by the MSLT. This study also sought to determine whether or not the increased sleep drive, due to the hypothesized sleep debt in sleepy normals, would attenuate the decreased sleep propensity typically associated with a change in the circadian timing of sleep.

10 8

6 4 9:30

11 :30

13:30

15:30

17:30

Hour

FIG. 1. Sleep latency (minutes) on each test for the moderately sleepy and alert normals (mean :!: SEM). Sleep. Va!. 19. No.5, 1996

The subjects were respondents to newspaper advertisements and announcements posted at local universities and hospitals soliciting healthy normal sleepers. In an extensive telephone interview the subjects reported their usual nocturnal bedtime, arising time, and total sleep time. All subjects were required to report spending between 6.5 and 8 hours in bed nightly and routinely going to bed between 2200 and 2400 hours (see Table 1). Subjects with highly irregular bedtimes and arising times (i.e. variations of >2 hours nightly) were excluded. The mean reported bedtime in the two groups did not differ: moderately sleepy = 10:57 (::f:6 minutes) p.m. and moderately alert = 10:51 (::±::14 minutes) p.m. These criteria were applied to assure that

PHASE ADVANCE

subjects not be obviously sleep deprived or have irregular sleep schedules. The daytime circadian pattern of screening sleep latencies on the MSLT (see Fig. 1) partially confirms the similarity of circadian rhythms between the study groups. Also excluded were subjects who reported difficulty sleeping at night, sleepiness during the daytime, or routinely napped. Subjects reported no history of alcohol or drug abuse and no current drug use. Each subject then came to the sleep center and underwent a medical history, drug use history, physical examination, and blood and urine samples were collected. Standard laboratory analyses of the blood and urine samples were used to verify normal health and the absence of recent drug use. The subjects also completed the Minnesota multiphasic personality inventory (MMPI), which was used to screen for the presence of any psychologic problems. Subjects underwent a standard screening 8-hour nocturnal polysomnogram (NPSG) and MSLT the following day. They reported to the sleep laboratory 1.5 hours before their usual bedtime with the latest bedtime allowed 2400 hours. These limits were established to maintain an 8-hour NPSG with an 0800 hours arising time the next morning and a minimum of 1.5 hours of wakefulness before the first latency test of the MSLT and also to standardize the timing of the sleep period for the 4-hour phase advance. For all the NPSGs subjects had electrodes attached at standard placements for the continuous recording of bilateral electrooculograms (EOG), submental electromyograms (EMG), central (C3/C4) and occipital (Oz) electroencephalograms (EEG), and electrocardiograms (V5) according to standard procedures (13). On the screening night only, nasooral thermistors were used to monitor breathing during sleep and tibialis EMGs to monitor leg movements (14). Subjects were required to have no respiratory disturbances during sleep (i.e. reduction of >50% in nasooral thermistor tracings) or leg movements during sleep on the screening NPSG. In addition, subjects were required to have a sleep efficiency of >85% (time asleep per time in bed) on the screening NPSG. In the morning subjects arose, bathed, and were allowed to eat a light breakfast with the instruction to avoid beverages with caffeine. Electrodes were checked and replaced if necessary to prepare for the MSLT. The MSLT was conducted at 930, 1130, 1330, 1530, and 1730 hours according to the standard MSLT procedures (15) with a previously demonstrated testretest reliability at this center of 90% (16). Subjects went to bed in a darkened room and were instructed to try to fall asleep, while EOGs, submental EMG, and EEGs, always including an Oz placement, were recorded. The recording was terminated after 1 minute

419

of unambiguous stage 1 sleep, the first signs of stage 2 or rapid eye movement (REM) sleep, or 20 minutes of continuous wake according to standard sleep stage criteria (13). Sleep latency was defined as minutes to the first 30-second epoch of nonwake recording. Results of the screening MSLT defined the two study groups (see Table 1 and Fig. 1). To qualify as moderately sleepy, subjects had average daily sleep latencies on the MSLT of :510 minutes and as moderately alert, subjects had latencies of 2":12 minutes. All subjects meeting the sleepy and alert criteria in a 2-month period were included in this study. This resulted in 11 moderately sleepy and 17 moderately alert subjects qualifying for the study. Within a week, but typically the following night, those qualified subjects spent two baseline nights in the sleep laboratory (8 hours bedtime) going to bed at their usual bedtime (2200-2400 hours). Then on night three each subject's bedtime was shifted by 4 hours in a phase advance (1800-2000 hours) with the time of arising 8 hours after bedtime. An 8-hour bedtime was enforced on all nights. Data collected on the second baseline laboratory night were used as the baseline night against which the night three shifted sleep was compared; the first night was considered an adaptation night. All sleep recordings were scored following the Rechtschaffen and Kales (13) criteria by scorers unaware of the study night. Scorers maintained an inter-rater reliability of 90% or better. The sleep parameters evaluated are listed in Table 2. The latencies to stage 1 or 2 are defined as minutes from lights out to the first epoch of that sleep stage and latency to persistent sleep is defined as minutes from lights out to the first epoch of 10 continuous minutes of nonwake recording. Wake during sleep is the minutes of wake time after persistent sleep is achieved and before the final awakening of the recording. Entries to wake or stage 1 are the number of shifts from other sleep stages to at least one epoch of wake or stage 1. Due to the large number of sleep parameters and their inherent correlation, multivariate analyses (MANOV A) were conducted using the general linear model analysis (SAS Institute) to compare the groups. The groups were compared on the baseline and the shift nights separately in two analyses of each night. Variables reflecting induction and maintenance of sleep were compared in one analysis and the sleep staging variables were compared in another analysis. A two-factor repeated measures design MANOVA conducted in each group separately was used to compare sleep efficiency on the two nights and hours of night. Finally, a univariate ANOVA, two factor mixed design was used to compare groups and nights (the repeated factor) on the sleep efficiency variable. Sleep, Vol. 19, No.5, 1996

420

T. ROEHRS ET AL. TABLE 2.

Sleep parameters on baseline and shift nights for study groups Moderately sleepy

Moderately alert

Base night

Shift night

Base night

Shift night

Sleep induction and continuity SIp efficiency Lat St I Lat St 2 Lat PS WDS Ent St I Ent wake

93.3 (2.9)** 10.0 (17.5) 12.5 (4.6)*** 14.5 (7.9)** 21.7 (15.4) 17.3 (10.6)** 21.7 (15.4)

88.1 12.3 16.3 24.0 30.8 17.3 17.1

(8.9)** (13.9) (13.2) (35.2) (20.6) (11.9)* (4.6)*

87.9 (5.7) 22.2 (10.7) 5.1 (11.3) 30.0(17.4) 32.7 (35.5) 7.6 (5.6) 22.9 (15.1)

69.3 46.2 49.4 51.7 78.6 10.4 13.1

Sleep stage measures 0/0 St 1 0/0 St 2 0/0 St 3/4 0/0 St REM

8.5 58.6 8.8 23.9

9.4 61.9 7.2 21.4

(4.6) (7.4) (6.5) (4.6)

6.3 57.3 15.7 20.5

10.9 (8.7) 59.0 (7.2) 12.5 (6.2) 17.4(7.8)

(3.4) (8.0) (7.3) (3.9)

(2.7) (6.1) (7.1) (4.6)

(21.3) (65.0) (64.9) (65.6) (98.2) (2.3) (10.2)

Data are means (::>:SD). Group differences * p < 0.05, ** P < 0.01, *** P < 0.001. Sip efficiency = % bedtime asleep; Lat St I or 2 = latency stage 1 or 2 sleep; Lat PS = latency to persistent sleep (10 minutes); WDS = wake during sleep; Ent St 1 or wake = entries from sleep to stage 1 or wake.

RESULTS The average daily sleep latency on the MSLT for the moderately sleepy group was 7.1 : :': : 2.1 and that of the moderately alert group was 16.8 : :': : 3.3 minutes (see Table 1). The pattern of the sleep latencies of each group throughout the day is illustrated in Fig. 1. By the screening criteria the groups were designed to differ significantly in average daily sleep latency (F = 101.71 P < 0.001). They were also selected to be similar in circadian phase, which is seen in the similar timing of both the nadir and peak in sleep latency. There was a significant time effect (F = 3.69 P < 0.007) but no significant group by time interaction (F = 0.15, NS). The shortest sleep latency in each group was recorded on the 1530 hours latency test, which differed from that at 1730 hours (F = 10.3, P < 0.003) and that at 930 hours (F = 6.29, P < 0.01). The sleep parameters for each group on the baseline and the shift nights are presented in Table 2. The MANOVA analyses of the sleep induction and continuity variables (see Table 2 for list of variables) on the baseline night showed significant overall group differences (F = 4.45, P < 0.004). Specific differences were found in the following variables: sleep efficiency (F == 8.46, P < 0.007), latency to stage 2 (F = 13.17, P < 0.001), latency to persistent sleep (F = 8.16, P < 0.008), and entries to stage 1 (F = 9.79, P < 0.004). Thus, on the baseline night the moderately sleepy group showed a faster sleep onset and less wakefulness than did the alert group. There were no group differences in the analysis of sleep stage variables (see Table 2 for the list). The analyses of the shift night data also revealed significant overall group differences in sleep induction and continuity variables (F = 3.04, P < 0.02). The groups differed specifically in sleep efficiency (F Sleep, Vol. 19, No.5, 1996

8.28, P < 0.007), entries to stage 1 (F = 5.51, P < 0.03), and entries to wake (F == 4.43, P < 0.04). Thus, on the shift night the moderately sleepy group compared to the moderately alert group showed less sustained wakefulness but more entries to stage 1 and wake, which were followed by a return to sleep. Again there were no significant group differences in the analysis of the sleep stage variables, To test for a differential response between groups to the sleep schedule shift, a mixed design, univariate ANOVA comparing the groups for sleep efficiency on the baseline and shift night was conducted. The analysis revealed significant between-group differences (F = 14.4, P < 0,001) as the two multivariate analyses above showed, It also indicated a significant effect of sleep schedule shift (between night difference) as previous sleep schedule shift studies have shown (F == 14.84, P < 0.001). Most important to the hypothesis of this study, a significant group by night interaction (F = 5.27, p < 0.03) was found. The moderately sleepy group was less affected by the sleep schedule shift than was the moderately alert group. The differential group response to the phase advance is best seen in comparing sleep efficiency between groups each hour of the night as is illustrated in Fig. 2. The moderately sleepy group showed a reduced sleep efficiency only for the first 2 hours of the shifted night (F = 4.19, P < 0.05), In contrast, the sleep efficiency of the moderately alert group was uniformly reduced on the shifted night relative to the baseline night and they differed significantly from the moderately sleepy group in the last 4 hours of the shifted night (F = 4.26, P < 0.02), DISCUSSION The present study reports normal sleep and phase advanced sleep differences between healthy normals

421

PHASE ADVANCE 100

TABLE 3.

Nocturnal sleep efficiency in normals with differing levels of daytime sleepiness Sleepya

~

~

Mean MSLT (minutes) Sleep efficiency (%TSTmB)

80

>u

s: ell '0

Moderately Moderately sleepy alert

Alerta

3.8 (L2)

7.1 (2.1)

16.8 (3.3)

18.0 (1.5)

94.6 (3.4)

93.3 (2.9)

87.9 (5.7)

86.9 (7.4)

Data are means (:!:SD). a Data from Roehrs et aL (6).

~

Cell ell

iii

60 Alert-Sase ........ Alert-Shft -B- Sipy-Sase --- Sipy-Shft -B-

40~r---r--~---'--~r---r---~--'----r---' o 4 6 7 8 9 2 3 5

Hour

FIG. 2. Sleep efficiency (%) for the moderately alert and sleepy normals on the baseline (Base) and shift (Shft) nights.

chosen for their differential levels of daytime sleepiness. These subjects reported 7-8 hours of regular nightly sleep and were without nocturnal sleep or daytime sleepiness complaints but showed relatively short versus long average daily sleep latencies on the MSLT. On the MSLT they showed identical temporal patterns in their daytime sleep latency suggesting that the groups had similar circadian phases. Thus, their differential response to the phase advance is likely due to differences in hypothesized sleep drives rather than basal circadian differences. At night these moderately sleepy subjects slept more efficiently than the moderately alert subjects but did not show reliable differences in sleep staging variables. As such, these data replicate the data of a previous study showing nocturnal sleep efficiency differences in normals who were chosen for even more extreme differences in level of daytime sleepiness than that of this study (7). The present data also extend the sleep efficiency findings to phase advanced nocturnal sleep (i.e. a 4-hour sleep schedule shift). Similar between-group differences in sleep efficiency were also seen on the phase advanced night. It could be argued that the sleepy subjects of this study are merely individuals who have the ability to fall asleep quickly and/or at unusual times of the day. Taken in isolation, this study alone certainly does not rule out such an explanation. However, the findings of a number of studies now converge to suggest that such individuals are mildly sleep restricted. That is, even though they report sleep times comparable to their

moderately alert counterparts, it is not sufficient because their biological sleep needs are somewhat greater. One variable in healthy normals that seems to reflect the habitual sleep insufficiency is sleep efficiency. It is interesting to compare the nocturnal sleep efficiency and daytime MSLT results of the earlier (7) and the pres&nt study (see Table 3). The moderately sleepy subjects in this study had longer average daily sleep latencies and lower sleep efficiencies than that of the sleepy subjects of the previous study. As well, the moderately alert subjects in this study had shorter average daily sleep latencies and higher nocturnal sleep efficiencies than the previously studied alert subjects. Between studies the nocturnal sleep efficiency seems to relate to average daily sleep latency. However, a single study comparing a range of daily sleep latencies and nocturnal sleep efficiencies is required to reach such a conclusion. Critical to the major hypothesis of this study, the moderately sleepy subjects showed reduced sleep efficiency only during the first part of the shifted sleep night. To some extent, the hypothesized sleep debt of the moderately sleepy subjects attenuated the disruptive effects of the phase advance. This is particularly true when compared to the alert subjects who had reduced sleep efficiency throughout the shift night. It is possible that an even sleepier group of subjects [e.g. average daily MSLT sleep latencies as in the earlier study (:::;6 minutes)] would have shown no disruption of sleep during this phase advance protocol (7). These data support notions regarding the interactions of homeostatic factors with circadian factors in determining the likelihood of the initiation and maintenance of sleep at different times of the day. The twoprocess model of Borbely and associates posits a homeostatic factor and a circadian factor that combine to predict the initiation and duration of sleep episodes (17). The model would predict that an elevated sleep drive can override to some degree circadian tendencies to alertness. The present data support such a view. However, there clearly are limits to which the homeostatic factor can override the circadian factor. Several investigators have noted that there is a circadian period during which sleep is not initiated and episodes of Sleep, Va!. 19, No.5, 1996

T. ROEHRS ET AL.

422

sleep are inevitably terminated. This period has been termed the "forbidden zone" (18). The phase advance in the present study extended into this circadian period of reduced sleep and both subject groups showed reduced sleep efficiencies during this period. Of particular interest is the fact that the alert subjects of this experiment experienced reduced sleep efficiency throughout the shifted sleep night. Initiating sleep between 1800 and 2000 hours had a disruptive effect even on their sleep during the period of time these subjects usually slept, from 2200 to 2400 hours and on. These data may then illustrate the potentially disruptive effects sleep in the evening (e.g. a nap) can have on nocturnal sleep, a frequently cited, but infrequently documented, sleep hygiene tenant. It should be noted that the full night reduction of sleep efficiency was only seen in the alert subjects. These present results can be discussed in the context of studies on jet lag. The 4-hour phase advance required in situ in this study is almost equivalent to a U.S. west coast to east coast flight or a U.S. east coast to England flight. The results of this study would suggest that a fully rested and alert person would experience a greater disruption of herlhis sleep than the more sleepy person, at least on the first night in the new time zone. And as noted above, the reduced sleep efficiency would be experienced throughout the sleep period. A jet lag study conducted by this center and a sleep center in England showed that the disrupted sleep associated with a flight to England from the U.S. eastern time zone (i.e. a 5-hour phase advance) can be improved with a short-acting benzodiazepine, which also leads to improved daytime alertness (12). In summary, these data are consistent with the notion that the daytime sleepiness of otherwise healthy normals is an expression of an accumulated sleep debt due to a chronic mild restriction of sleep time relative to biological sleep need. Given the presence of a heightened sleep drive, we predicted these moderately sleepy normals, compared to their alert counterparts, would respond differently to a sleep phase advance. A sleep phase advance is associated with reduced sleep propensity and we hypothesized that the heightened sleep drive would attenuate the reduced sleep propensity. The data showed just that. Of course, the direct test of the hypothesis that the sleepiness of healthy normals with unusually short MSLT scores is an expressed chronic sleep debt is to extend their sleep time and show improved MSLT scores. An earlier study did find improved MSLT and performance scores in sleepy normals after six consecutive nights of 10 hours time in bed (5), although the sleepy subjects never achieved the level of alertness seen in the alert subjects of that study. The follow-up study in which the period of Sleep, Vol.

19,

No.5, 1996

sleep extension was increased to 14 consecutive nights did show further improvements in MSLT scores (6). But that study also found that not all the subjects benefited from the sleep extension. Some never increased their MSLT scores beyond a daily mean of 8 minutes. Why this subgroup did not benefit remains to be determined. Acknowledgements: This work was supported by National Institutes of Health NIAAA grant ROI-AA07147, NIDA grant ROI-DA05086, and NHLB grant P50HL42215.

REFERENCES 1. Levine B, Roehrs T, Zorick F, Roth T. Daytime sleepiness in young adults. Sleep 1988;11:39-46. 2. Roth T, Roehrs T, Carskadon M, Dement W: Daytime sleepiness and alertness. In: Kryger M, Roth T, Dement W, eds. Principles and practice of sleep medicine. Philadelphia: WB Saunders, 1989:14-23. 3. Webb WB, Agnew HW. Are we chronically sleep deprived? Bull Psychon Soc 1975;6:47-8. 4. Carskadon MA, Mancuso J, Keenan S, Littel W, Dement WC. Sleepiness following oversleeping. Sleep Res 1986;15:70. 5. Roehrs T, Timms V, Zwyghuizen-Doorenbos A, Roth T. Sleep extension in sleepy and alert normals. Sleep 1989;12:449-57. 6. Roehrs TA, Shore E, Papineau K, Rosenthal L, Roth T. A two week sleep extension in sleepy normals. Sleep Res 1994;23: 142. 7. Roehrs T, Timms V, Zwyghuizen-Doorenbos A, Buzenski R, Roth T. Polysomnographic, performance, and personality differences of sleepy and alert normals. Sleep 1990;13:395-402. 8. Rosenthal L, Kontich D, Folkerts, Fortier J, Roehrs T, Roth T. Total sleep time during an enforced time in bed period in sleepy and alert subjects. Sleep Res 1993;22:99. 9. Rosenthal L, Roehrs TA, Krstevska S, Rosen A, Sicklesteel J, Roth T. Auditory awakening thresholds in sleepy, alert, and sleep deprived subjects. Sleep Res 1992;21:111. 10. Seidel WF, Roth T, Roehrs T, Zorick F, Dement WC. Treatment of a 12-hour shift of sleep schedule with benzodiazepines. Science 1984;224: 1262-4. II. Dinges D. The influence of the human circadian timekeeping system on sleep. In: Kryger M, Roth T, Dement W, eds. Principles and practice of sleep medicine. Philadelphia: WB Saunders, 1989: 153-62. 12. Nicholson AN, Pascoe PA, Spencer MB, Stone BM, Roehrs T, Roth T. Sleep after transmeridian flights. Lancet 1986;Nov: 1205-8. 13. Rechtschaffen A, Kales A. A manual of standardized, techniques and scoring system for sleep stages of human sleep. Los Angeles, CA: Brain Information ServicelBrain Research Institute, University of California at Los Angeles, 1968. 14. Bornstein SK. Respiratory monitoring during sleep: polysomnography. In: Guilleminault C, ed. Sleeping and waking disorders: indications and techniques. Menlo Park, CA: AddisonWesley, 1982. 15. Carskadon MA, Dement WC, Mitler MM. Guidelines for the multiple sleep latency test (MSLT): a standard measure of sleepiness. Sleep 1986;9:519-24. 16. Zwyghuizen-Doorenbos A, Roehrs T, Schaefer M, Roth T. Testretest reliability of the MSLT. Sleep 1988;11:562-5. 17. Daan S, Beersma DGM, Borbely AA. Timing of human sleep: recovery process gated by a circadian pacemaker. Am J Physioi 1984;246:RI61-83. 18. Lavie P. Ultrashort sleep-waking schedule. III. "Gates" and "forebidden zones" for sleep. Electroencephalogr Clin Neurophysiol 1986;63:414-25.