WRIST ACTIGRAPHIC MEASURES OF SLEEP IN SPACE
Wrist Actigraphic Measures of Sleep in Space
Timothy H. Monk, DSc; Daniel J. Buysse, MD; and Lynda R. Rose
Sleep and Chronobiology Center, University of Pittsburgh Medical Center, Western Psychiatric Institute and Clinic
3811 O'Hara St, Pittsburgh PA 15213 USA
Study Objectives: To determine whether wrist actigraphy is useful as a tool for space-based sleep research. Specifically, to determine whether bedtimes and waketimes can be identified from the actigraphic record, and whether actigraphic measures of sleep in
space are related to polysomnographic (PSG) ones.
Design and Setting: Actigraphy, sleep diary, and Polysomnographic (PSG) measures of sleep were obtained from four subjects in
two 72h measurement blocks occurring 2d and 12d into a 17d Space Shuttle mission in orbiting the earth in microgravity.
Patients: Four healthy male astronauts aged 38y – 47y.
Interventions: NA
Measurements and Results: Sleep onset and offset at "night" could be quite clearly identified from the actigraphic record and were
better estimated by actigraph than by diary. There was a high correlation between actigraphic and PSG estimates of sleep duration (r=0.96) and sleep efficiency (r=0.88), and a similarity in the mean estimates obtained. On a minute-by-minute basis, there was
a good correlation between sleep stage and actigraphic movement counts, with a higher level of counts per minute recorded in
epochs with lighter PSG sleep stages. There was also a high correlation (r=0.90) between minutes of stage 0 (wake) occurring
between bedtime and wake time, and number of non-zero actigraph epochs during the same interval.
Conclusions: Actigraphy worked well in space both as a way of detecting bedtimes and waketimes, and as an indicant of sleep
restlessness.
Key words: Human sleep; rest; activity; actigraph recording; space; microgravity; astronauts
INTRODUCTION
ant, others miss the proprioceptive cues of the head on a
pillow, and without doubt, most space ships are noisy,
occasionally have troubling drafts, and are sometimes too
hot or cold for comfort. There are also likely to be sleep
disruptions due to circadian dysfunction, especially if the
mission requires around-the-clock coverage, or if abrupt
shifts in routine need to be made for operational reasons. In
a recent publication5 we report that such circadian dysfunction can be avoided if the mission is appropriately scheduled (i.e., that microgravity and/or space travel per se need
not themselves be a source of circadian desynchrony.)
Sleep diary reports from 24 subject-nights in flight indicated the following sources of sleep disruption: (reported as
number of subject-nights): ambient temperature (6/24),
noise (3/24), need to void (3/24), vibration (1/24), mission
circumstances (1/24), and general discomfort (1/24). On
4/24 subject nights a sleep disruption was reported as "just
woke" (i.e., of unknown etiology).
Because sleep in space is so important, it behooves the
space life science researcher to include sleep measures in
studies of astronauts in space. However, the polysomnographic (PSG) study of sleep in space is enormously expensive in terms of both money and astronaut time and effort.
Since the pioneering studies of Frost and colleagues during
the 1970s Skylab missions,6 the only published PSG space
BEFORE HUMAN BEINGS WENT INTO SPACE their
was some doubt about whether they would be able to sleep
at all in microgravity.1,2 These doubts were soon allayed,
but as Santy3 remarks in her survey study of Space Shuttle
astronauts, sleep remains a significant problem for many
astronauts, particularly those working an around-the-clock
"dual-shift" mission, where sleeping pill use can average
50%. There are a number of reasons why people may not
sleep as much in space as on the ground. First of all, there
is the excitement and stress of the flight, which often represents the chance of a lifetime and the culmination of
many years of preparation. Anecdotally, some sleep
restriction in space is voluntary, so that Earth viewing can
occur, since the astronauts' schedule is often so busy that
few other opportunities for such viewing arise. Early on in
the flight there may be disruptions due to space adaptation
sickness, though these tend to fade after a few days.4
Although many report sleeping in microgravity to be pleasAccepted for publication April 1999
Address correspondence to: Timothy H. Monk, D.Sc., Sleep and
Chronobiology Center, University of Pittsburgh Medical Center,
Western Psychiatric Institute and Clinic, 3811 O'Hara St, Pittsburgh PA
15213 USA, Tel:412-624-2246. Fax: 412-624-2841. E-mail:
[email protected]
SLEEP, Vol. 22, No. 7, 1999
948
Wrist Actigraphic Measures of Sleep in Space—Monk et al
Subjects
studies have been the investigations of Gundel7,8 aboard
Space Station Mir, and our own aboard the Space Shuttle
Columbia.5 Clearly, wrist actigraphic measures would be a
much more acceptable alternative in terms of cost and time
line planning. However, in the absence of gravity, the individual is essentially floating about in both sleeping and
waking states, and the question arises as to whether actigraphy in space works as well as it does on the ground.
Because of the extreme time pressure faced by astronauts,
actigraphy is preferable even to simple diary measures, let
alone PSG ones. Thus, the present study had two main
research questions: 1) Can actigraphy be used in space as a
method of differentiating sleep from wakefulness, and thus
to estimate bedtime, waketime, sleep efficiency, and sleep
duration? and 2) How well do actigraphic estimates of
sleep in space relate to PSG ones (e.g., in relation to sleep
stages and wakefulness during the "night")? The present
brief report compares actigraphic and PSG measures of
sleep from four astronauts who each gave six recorded
nights of sleep in flight aboard a 1996 Space Shuttle mission. Other aspects of that experiment have been reported
elsewhere.5
The subjects were four of the seven astronauts aboard
the mission. All were males, ages ranged from 38years of
age to 47years of age (mean=42.5years of age). Two of the
subjects were members of the NASA astronaut corps., and
two were payload specialists recruited specifically for this
mission. All were extremely healthy and reported no sleep
abnormalities. Arbitrary subject code numbers (S1 through
S4) will be used throughout this report. The study was
approved by both University of Pittsburgh and NASA
Internal Review Boards for the ethical treatment of human
subjects. The experiment conformed to the Declaration of
Helsinki.
Measurement Block
The experiment was structured in terms of 72hours
measurement blocks within which all sleep and circadian
rhythm measures were taken. The two in-flight measurement blocks occurred towards the beginning "early flight"
(starting 2days after launch) and towards the end "late
flight" (starting 12days after launch) of the 17-day mission.
Each measurement block ran for 72-hours, starting and
ending at the beginning of the workday. For subject S3, the
early flight measurement block started approximately
12hours late, but then ran for a full 72-hours.
METHODS
The Space Mission
This experiment comprised experiment E-948 of the 17day Life and Microgravity Studies (LMS) Spacelab mission that flew aboard Space Shuttle Columbia as STS-78.
Lift-off was at 10:50 EDT on June 20, 1996, recovery at
08:38 EDT on July 7, 1996. Subjects slept in individual
bunks in the mid-deck of the spacecraft which provided
shielding from light and noise. Most of the working day
was spent in the Space Shuttle module (volume 32 cubic
meters) which was of dim illumination, (<200 Lux).
Ambient temperatures in the spacecraft averaged about
23°C.
Strenuous efforts were made to ensure that the timing
and trajectory of the mission disrupted the astronauts' circadian rhythms as little as possible. Lift-off was in daylight
(10:50 EDT), requiring only a slightly earlier than usual
waketime (05:50 EDT). For operational reasons, there
were changes in bedtime and waketime (and thus the whole
routine) throughout the flight, but these were distributed
equally throughout the mission leading to a phase advance
of 25 minutes per day for all but one day of the mission.
Strict rules (denoted by NASA "Appendix K") governed
the astronaut's time availability for work. An eight-hour
sleep opportunity was mandated for each day, as well as
rest times at either end of the sleep period and time off for
meals. For the present experiment, time was set aside during the work shift to "pay back" for time spent in EEG
wiring, etc., close to the sleep episode.
SLEEP, Vol. 22, No. 7, 1999
Polysomnographic (PSG) Sleep Recording
Each sleep episode during the 72-hours measurement
block was recorded using PSG. Electrodes recording EEG,
EOG, and EMG were positioned on the subject's head each
"night" of study using a cap-based system with both builtin and disposable electrodes. EEG electrodes were placed
in the C3 and C4 positions, which were referenced to A1
and A2 mastoid electrodes. EOG was measured by electrodes asymmetrically placed at the outer canthi, EMG by
submental electrodes. Two ground electrodes were placed
on the forehead. Impedances were required to be 10 KOhms or better. Signals were recorded onto magnetic tape
using Oxford Medilog© recorders. After recovery, the
tapes were played back in order to visually score each
minute of sleep into conventionally defined sleep stages (0,
1, 2, 3, 4 and REM), according to standard criteria.9
Computerized analysis of the PSG record was also undertaken, but will not be reported here. The recorders were
initialized several months before launch; no further adjustments to the devices were made.
Subjective Sleep Diary
After each sleep episode during the 72-hours measurement block, each subject completed a modified computerbased version of the Pittsburgh Sleep Diary (PghSD).10
This brief instrument assessed the estimated timings of
949
Wrist Actigraphic Measures of Sleep in Space—Monk et al
Figure 1. Astronaut subject S1: 72-hour actigraph record, each point represents the mean activity count from five one-minute epochs. Time is Greenwich Mean
Time. The times at which the subject went to bed, first attempted sleep and finally got up (from the post-sleep diary) are also marked on the record (see key).
were made.
sleep onset and offset, recorded the number and nature of
within-night awakenings, the use of space motion sickness
and hypnotic medications, and concluded with visual analog scale ratings of sleep quality, mood, and alertness.
Diary acquired bedtimes and waketimes were used as
approximate guides for the application of the actigraphybased sleep onset and offset detection algorithms and thus
for the definition of "night."
RESULTS
From both post-flight interview and sleep diaries it was
clear that no medications were used which might have
influenced sleep. Figures 1 through 4 illustrate the 72-hour
actigraphic records (wrist activity) for subjects 1 through 4,
respectively. Also included in the graphs are the times at
which subjects went to bed, attempted sleep, and finally
awoke, as recorded on the post-sleep diary. Mean actigraphic counts per hour averaged across the four subjects
and two 72h measurement blocks were 83 counts/hour
between (diary defined) bedtime and waketime, and 891
counts/hour between waketime and bedtime (t=24.1, df=6,
p<0.0001). Thus, at least by simple inspection, it would
appear that actigraphy "works" in weightlessness, and that
the record can be used to identify approximate bedtimes
and wake times. The light level record from the Actillume©
also appeared to give a good indication of when the astro-
Wrist Actigraphy
For the duration of each 72-hour measurement block,
each subject wore an Actillume© on the non-dominant arm.
Actigraphy involves wearing an electronic device which
uses a piezo-electric crystal to count and then record the
number of arm movements per epoch (usually one minute).
In the Actillume© device, this is supplemented by the
simultaneous recording of ambient light levels (in this case,
as measured at the wrist). Actillumes© were initialized two
weeks before launch; no further adjustments to the devices
SLEEP, Vol. 22, No. 7, 1999
950
Wrist Actigraphic Measures of Sleep in Space—Monk et al
Figure 2. Astronaut subject S2: 72-hour actigraph record, each point represents the mean activity count from five one-minute epochs. Time is Greenwich Mean
Time. The times at which the subject went to bed, first attempted sleep and finally got up (from the post-sleep diary) are also marked on the record (see key).
nauts retired to their light-shielded "bunks."
Because the dynamics of body movement are very different in microgravity, we used a very simple algorithm to
detect sleeps, avoiding the more sophisticated algorithms
which are based on actual number of counts.11 Instead, a
simple algorithm of the beginning of five consecutive
epochs of zero counts (around the known approximate bedtime) was used to define actigraphic sleep onset, and the
beginning of a run of five consecutive non-zero epochs
(around the known approximate wake time) used to define
sleep offset. When measures of sleep onset and offset were
thus gleaned from the actigraphic record, and compared to
PSG estimates, quite good agreement was obtained on
nearly all of the 21 in-flight sleep episodes for which both
measures were available. In one "subject-night" the actigraphy algorithm failed to detect a very long (98 min.) sleep
onset latency, and thus grossly mis-estimated sleep onset.
For the other 20 "subject-nights," 20% of actigraphy sleep
onset estimates were within five minutes of the PSG estimate, 75% were within 15 minutes. Estimates of sleep offSLEEP, Vol. 22, No. 7, 1999
set were even better, of the 20 "subject-nights," 50% were
within five minutes of the PSG estimate (last epoch of actual sleep), 95% were within 15 minutes. Using these figures
alone (i.e., not accounting for WASO), mean sleep duration
estimates were 6.24h (+/- 0.60h s.d.) from actigraphy versus 6.14h (+/- 0.50h s.d.) from PSG. In 17 of the 20 subject-nights duration was (slightly) overestimated by the
actigraphy algorithm relative to PSG (t=2.8, df=18,
p<0.05). The correlation between the actigraphic estimates
and the PSG estimates of sleep duration for the 20 "subjectnights" was 0.969 (p<0.0001).
Again using the 20 subject-nights for analysis, a comparison was made between actigraph and diary estimates of
the times of initial sleep onset (at the beginning of the
"night") and final sleep offset (at the end of the "night"),
using PSG derived estimates as the "gold standard." The
mean absolute difference (minutes) from the PSG derived
time was smaller for actigraph than for diary estimates,
both for initial sleep onset (diary: 33.7 ± 30.5; actigraph:
11.5 ± 6.4; t(18) = 3.29, p< 0.01), and for final sleep offset
951
Wrist Actigraphic Measures of Sleep in Space—Monk et al
Figure 3. Astronaut subject S3: 72-hour actigraph record, each point represents the mean activity count from five one-minute epochs. Time is Greenwich Mean
Time. The times at which the subject went to bed, first attempted sleep and finally got up (from the post-sleep diary) are also marked on the record (see key).
(diary: 12.4 ± 10.7; actigraph: 6.3 ± 5.2; t(18) = 2.33, p<
0.05). Thus, even using a fairly unsophisticated algorithm,
actigraphic estimates appeared to be more accurate than
diary-based ones.
Comparison of minute-by-minute actigraphic and PSG
records was also undertaken in order to determine whether
actigraphy was also useful within the sleep episode. Using
the 20 "subject-nights" analyzed above, the mean actigraphic count (counts per hour - cph) for each 60-second
epoch of sleep was calculated, and then categorized by the
sleep stage it was scored into from the PSG record. The
mean overall actigraph cph was then calculated for each
stage of sleep. As one would predict, stage 0 (wake) was
associated with the highest level of activity (290 cph).
Importantly, though, the other stages of sleep showed actigraph counts that were intuitively related to the "depth" of
PSG-scored sleep. Thus stage 1 was the next highest (50
cph), followed by stage 2 (10 cph), with stage REM intermediate (27 cph). Delta sleep (stages 3 and 4) showed the
lowest levels of activity (8 cph). All four subjects exhibitSLEEP, Vol. 22, No. 7, 1999
ed this overall pattern of results. As a test of whether actigraphy could additionally be used as an overall measure of
sleep restlessness, the minutes of wakefulness (stage 0)
between bedtime and wake time from PSG for each subject-night was correlated with the number of non-zero
epoch counts (in the same interval) from the actigraph
record. A high correlation emerged (r=0.904, df=18,
p<0.0001). A scattergram of that relationship is plotted in
Figure 5. Thus, sleep restlessness as measured by the PSG
was associated with restlessness as measured by the actigraph.
As a determination of whether actigraphy could be used
to provide a numerical estimate of sleep efficiency, the percentage of non-zero epochs between bedtime and waketime
was calculated for each of the 20 subject-nights and compared with PSG measured sleep efficiency. Surprisingly, in
view of the simplicity of algorithm used, the two estimates
of sleep efficiency were very similar (Actigraph: 88.6%±
6.4%; PSG: 89.5%±6.1%), and were highly correlated (r =
0.882, p<0.001).
952
Wrist Actigraphic Measures of Sleep in Space—Monk et al
Figure 4. Astronaut subject S4: 72h actigraph record, each point represents the mean activity count from five one-minute epochs. Time is Greenwich Mean Time.
The times at which the subject went to bed, first attempted sleep and finally got up (from the post-sleep diary) are also marked on the record (see key).
DISCUSSION
space should be recorded. While this manuscript was being
prepared, the STS-90 "Neurolab" mission had just returned
to Earth with an experiment of Czeisler, Dijk, and colleagues that had used more advanced electronic techniques
to acquire PSG records in space with apparent success.
Recent Space Station Mir experiments by Moldofsky and
by Hobson & Stickgold have also included nights of sleep
recording by PSG and by the Nightcap© systems respectively. However, the extreme convenience of actigraphy
does render it eminently suitable for situations in which life
science experiments are not the main focus of the space
mission. As shown by the present study, when so used, one
can get a fairly accurate determination of bedtime, wake
time, and sleep duration. It should be noted, though, that in
one case our simple actigraphy algorithm failed to detect a
long (98 min.) sleep onset latency, a finding which has been
noted in previous actigraph-PSG comparisons.11 Although
the above estimates did not account for WASO, there was a
strong relationship between cumulative duration of wake
episodes between bedtime and waketime and number on
Clearly, no one would deny that should it be available,
PSG ought to be the "gold standard" by which sleep in
Figure 5. Scattergram of 20 subject-nights (see text) plotting minutes of
wake during the sleep episode against number of non-zero actigraph counts
during the same interval of time.
SLEEP, Vol. 22, No. 7, 1999
953
Wrist Actigraphic Measures of Sleep in Space—Monk et al
non-zero actigraph counts, as well as a meaningful minuteby-minute relationship between actigraph counts per hour
and PSG stage of sleep. Indeed, actigraph and PSG estimates of sleep efficiency were within one percentage point
of each other. Thus, with caution, actigraphy may in some
case be a useful technique for studying sleep in space when
PSG is infeasible.
ACKNOWLEDGMENTS
We thank the astronauts for doing the study, Bart Billy,
Kathy Kennedy and Linda Willrich for seeing the project
through to completion, Timothy Hoffman for computer
programming, our many NASA colleagues for their support, Jan Matzzie and Jim Monahan for sleep scoring, and
Jean Miewald for data analysis help. Primary financial support was provided by NASA contract NAS 9-18404, additional support was provided by NAS 9-19407, MH 01235,
AG 15136 and AG 13396. Data processing support was
provided by MH 30915 (PI David Kupfer, M.D.).
REFERENCES
1. Graeber RC. Sleep in space. In: NASA Ames Research Center, ed.
Sleep and it's applications for the military (Proceedings of the XXVII
NATO DRG Seminar). Moffett Field, CA: Aerospace Human Factors
Research Division, 1987:16-17.
2. Stampi C. Sleep and circadian rhythms in space. J Clin Pharmacol
1994;34:518-534.
3. Santy PA. Analysis of sleep on shuttle missions. Aviation, Space and
Environmental Medicine 1988:59:1094-1097.
4. Nicogossian AE, Huntoon CL, Pool SL. Nicogossian AE, Huntoon CL
and Pool SL, eds. Space physiology and medicine. Philadelphia, PA:
Lea & Febiger, 1989.2:
5. Monk TH, Buysse DJ, Billy BD, Kennedy KS, Willrich LM. Sleep
and circadian rhythms in four orbiting astronauts. J Biol Rhythms
1998;13:188-201.
6. Frost JD, Shumate WH, Salamy JG, Booher CR. Sleep monitoring: the
second manned Skylab mission. Aviat Space Environ Med
1976;47(4):372-382.
7. Gundel A, Nalishiti V, Reucher E, Vejvoda M, Zulley J. Sleep and circadian rhythm during a short space mission. Clin Investig
1994;71(9):718-724.
8. Gundel A, Polyakov VV, Zulley J. The alteration of human sleep and
circadian rhythms during spaceflight. J Sleep Res 1997;6:1-8.
9. Rechtschaffen A, Kales AA. A manual of standardized terminology,
techniques, and scoring system for sleep stages of human subjects.
Bethesda, MD: National Institute of Neurological Diseases and
Blindness, 1968.
10. Monk TH, Reynolds CF, Kupfer DJ, et al. The Pittsburgh sleep diary
(PghSD). J Sleep Res 1994;3:111-120.
11. Sadeh A, Sharkey KM, Carskadon MA. Activity-based sleep-wake
identification: an empirical test of methodological issues. Sleep
1994;17:201-207.
SLEEP, Vol. 22, No. 7, 1999
954
Wrist Actigraphic Measures of Sleep in Space—Monk et al