Sleep, 18(1):11-21
© 1995 American Sleep Disorders Association and Sleep Research Society
Shift Work and Sleep Deprivation
Improving Adaptation to Simulated Night Shift:
Timed Exposure to Bright Light Versus Daytime
Melatonin Administration
Drew Dawson, Nicola Encel and Kurt Lushington
Department of Obstetrics and Gynaecology, The Queen Elizabeth Hospital University of Adelaide.
South Australia
Summary: Chronic circadian disturbance is thought to cause many of the health and social problems reported by
shift workers. In recent years, appropriately timed exposure to bright light and exogenous melatonin have been
used to accelerate adaptation to phase shifts of the circadian system. In this study we compared adaptation to night
shift in three groups of subjects. The first treatment group received timed exposure to bright light (4-7,000 lux
between 2400 and 0400 hours on each of three night shifts). The second treatment group received exogenous
melatonin by capsule (2 mg at 0800 hours then I mg at 1100 and 1400 hours). The placebo control groups received
either dim red light at less than 50 lux or placebo (sucrose) in identical capsules at the same time. Results indicated
that all groups shifted significantly from baseline. Using the dim-light melatonin onset as a circadian marker, the
bright-light group shifted the furthest, whereas there was no significant difference between the melatonin and placebo
groups. Sleep quality as determined by wrist actigraphy was most improved in the light-treatment group, although
the melatonin group also showed significant improvements. Cognitive psychomotor performance was most improved
in the light-treatment group and the melatonin group again showed little difference from the control group. Although
melatonin was unable to increase the amount of the phase shift following transition to night shift, it -is likely that
the intermediate levels of improvement in sleep reflect the hypothermic effects of melatonin. By lowering core
temperature across the sleep period, sleep may be enhanced. This improvement in sleep quality did not produce
concomitant improvements in shift performance for the melatonin group. This suggests that the enhanced performance in the light-treatment group may reflect more direct "energizing" effects. On the basis of these results, bright
light is clearly superior in its ability to phase shift the circadian system and thereby improve sleep and performance.
However, melatonin may permit shift workers to override the circadian system for short periods and avoid the
potential toxicity due to overzealous manipUlations of the circadian pacemaker. In rapidly rotating shift schedules,
melatonin may be preferable because it would not require workers to reverse the large phase shift induced by light.
Key Words: Light-Melatonin-Biological rhythms-Shift work-Sleep-Cognitive pefformance-Core temperature.
Circadian disturbance is thought to be responsible
for many of the medical and psychological problems
associated with shift work (1). In recent years, both
behavioral and physiological approaches have been
employed to facilitate adaptation to shift work (for a
comprehensive review of this topic see reference 2).
These treatment strategies are based on the notion of
accelerating the rate at which the circadian system can
adapt to a phase shift of the circadian system, thereby
reducing the degree to which workers are required to
sleep and work at inappropriate phases of the circadian
cycle.
By reducing the degree of circadian de synchrony,
studies of simulated transitions to night shift have
shown that accelerated circadian adaptation is associated with significant improvements in day sleep and
nighttime alertness (3-5). On the basis of these studies,
it has been hypothesized that increasing the rate of
circadian adaptation may help attenuate the long-term
health problems associated with rotating shift work.
Using appropriately timed exposure to bright light,
it is possible to rapidly shift the phase of the circadian
system (6-10), thereby increasing the normal rate of
circadian adaptation (3-5). By rapidly phase delaying
the circadian system it is possible to reestablish the
Accepted for publication July 1994.
Address correspondence and reprint requests to Dr. Drew Dawson, Department of Obstetrics and Gynaecology, The Queen Elizabeth Hospital, 11-23 Woodville Rd, Adelaide 5011, Australia.
11
.............-----------------------12
D. DA WSON ET AL.
normal phase angle between the circadian and sleepwake cycles. Returning the circadian and sleep-wake
cycles to their normal entrained position is thought to
reduce the degree of sleep disruption (11). This in turn
has been shown to improve alertness (4) and cognitive
psychomotor performance (3) at night.
Similarly, melatonin has been shown to shift the
phase ofthe circadian system (12,13). The phase-shifting effects of melatonin, however, appear to be the 180°
out of phase with those reported for light (12).
On the basis of published human phase-response
curves for bright light (7) and melatonin (12), early
morning light is thought to phase advance, whereas
morning melatonin phase delays. Conversely, evening
light phase delays and evening melatonin phase advances the circadian system. To accelerate adaptation
to a typical phase shift experienced during transition
to night shift, it is reasonable to suggest that workers
should delay their circadian system through appropriately timed bright light and melatonin. Melatonin,
however, may have a hypnotic effect independent of
its circadian effect (14-18). This hypnotic effect may
be related to a hypothermic effect of melatonin administration (19-22), and it has been shown that lowering core body temperature can facilitate sleep (23).
If this is the case, melatonin may have both circadian
and hypnotic effects that may be useful in the treatment
of shift work-related sleep disorders.
Although a combined strategy of bright light and
melatonin would appear reasonable, it is important to
compare the relative effects of each treatment on its
own. In this study, we compared the efficacy of nocturnal bright light and daytime administration of melatonin with a placebo group to improve sleep and performance measures during a 3-day transition to night
shift.
control were run between March and May 1992. The
melatonin group and its control were run between August and September 1992.
Protocol
The experimental protocol is outlined in Fig. 1. Subjects entered the laboratory during the evening preceding the 1st day of the study and went to bed between
2400 and 0800 hours the following morning. On rising,
subjects washed, showered and had breakfast between
0800 and 0900 hours, then spent the day (0900-1700
hours) training on the work and performance tasks
used in the study. At approximately 1730 hours, subjects had an indwelling venous catheter placed in the
antecubital vein and 5-ml blood samples were drawn
at I-hour intervals until 0200 hours. Subjects then
went to bed and were awakened at 0900 hours. Subjects
then commenced a second training session on the work
and performance tasks (1100-1700 hours). At 1730
hours they were cannulated on the opposite arm to the
previous night and 5-ml blood samples were drawn
hourly until 0200 hours.
Subjects commenced the night shift between 2300
and 0700 hours. During each of the three night shifts,
subjects did performance tests hourly. Between 0400
and 0700 hours subjects worked on a variety of clerical
tasks designed to simulate a typical shift-work load.
Subjects had a III hour break after the performance test
at 0400 hours. Following each of the three night shifts,
subjects had 2 hours of free time in which they could
eat breakfast and prepare for sleep. During this time
subjects were kept inside and not exposed to lighting
levels greater than 50 lux. Subjects retired at 0900
hours and were required to remain in bed until 1700
hours in the afternoon. On waking, subjects were permitted free time. During this time they were allowed
to go outside.
METHODS
Following the third day's sleep, subjects were cannulated
at 1730 hours in the same manner as the first
Subjects
2 nights of the study. Five-milliliter blood.<samples
Thirty-six male and female subjects between 18 and were drawn hourly for 24 hours.
30 years of age (mean age 23.6, SD = 3.9) gave informed consent to participate in this study. This project was part of a larger research program investigating Treatments
the effects of shift work on sleep and performance in
shift workers. This particular project was carried out
The bright-light group (n = 8) received a 4-hour
between March and September 1992 in Adelaide, South exposure to bright artificial light (average intensity
Australia (latitude 35°S). Groups of three or four sub- 4,000-7,000 lux) between 2400 and 0400 hours on
jects were required to spend 6 days in the Sleep and each of the 3 nights of the study. Compliance was
Circadian Rhythms Laboratory at The Queen Eliza- ensured by having subjects watch a video monitor with
beth Hospital. Subjects were randomly allocated to one bright lights mounted on either side and above (Apollo
offour groups: a bright light group (n = 8), a melatonin light boxes, Apollo Corp., UT, U.S.A.) throughout the
group (n = 12) and a placebo control group for each exposure period. The placebo group for the bright-light
condition (n = 8, 8). The light group and its placebo condition (n = 8) were exposed to dim red light at 50
Sleep, Vol. 18, No.1, 1995
13
BRIGHT LIGHT VERSUS MELATONIN
Adaptation
Training
Baseline
Training
Baseline
Segment
1
Shift 1
Day 1
Shift 2
Day 2
Shift
transition
segment
Shift 3
Day 3
~:~IIIIIIIIIIIIIIIIII~:~IIIIIIIIIIIIIIIIII~:~IIIIIIIIIIIIIII11II1~:llIIIIIIIIIIIIIIE~llllllIIllIIlllllllc:illlllllllllllllll~1111IIIIIIIIIIIIIII~IIIIIIIIIIIIIII~:~
23
24
02
01
03
04
05
23
Sleep
07
---
Post transition
evaluation segment
Day 4
II
06
111
8
Leisure time
III
16
Work task
20
24
~:::~ Performance test
FIG. 1. Schematic diagram of the experimental design. Subjects underwent a 6-day protocol divided into three segments: a baseline
segment in which they entered the laboratory at 0800 hours on the first day and underwent training on performance tests and circadian
phase assessment, a shift transition segment where subjects worked for 3 nights between 2300 and 0700 hours and slept between 0900 and
1700 hours, and then a posttransition segment when their circadian phase was assessed with a 24-hour melatonin profile. The inset between
day 3 and 4 shows an enlargement of the shift period where subjects underwent performance measures hourly across the night.
lux under identical viewing conditions as the treatment
group.
The melatonin group (n = 12) were administered
oral melatonin in gelatin capsules. To maintain plasma
melatonin levels at or above physiological levels for
the entire sleep period, the melatonin was divided into
three separate doses. Subjects in the melatonin group
received 2 mg at 0800 hours, 1 mg at 1100 hours and
1 mg at 1400 hours. As a placebo control for the melatonin treatment, the second placebo group (n = 8) received identical placebo medications at the same times
as the melatonin group.
Measures
Circadian phase
The initial phase position of the circadian system
for each subject was inferred from the phase of the
melatonin rhythm on the 1st and 2nd night. The final
phase position was inferred from the phase of the onset
of the melatonin rhythm measured on the last day of
the study. The onset of the plasma melatonin rhythm
was defined as the time of the first consistent level
above assay sensitivity according to the method outlined by Lewy and Sack (24). Samples were assayed
according to a previously described method (25). Intraand interassay coefficients of variation were 10% and
13%, respectively.
Core temperature
Body temperature was monitored in all subjects
throughout the 6-day study. Temperature was sampled
at I-minute intervals using indwelling rectal probes
(YSI-4400) inserted to a depth of lO cm connected to
a Vitalog PMS-8 ambulatory recording system. The
data were averaged into 60-minute bins across each of
the 3-day sleep periods for each group. Mean temperature across each sleep period was compared using a
two-factor repeated-measures ANOVA. The betweensubjects factor was group (bright light, melatonin and
Sleep. Vol. 18. No. I. 1995
14
D. DAWSON ET AL.
placebo) and the within-subjects factor was time (day
sleep 1, 2 and 3).
period in one of the subjects in the melatonin group,
which was lost due to a technical problem.
Sleep quality
Cognitive performance
Sleep quality was estimated from activity records
recorded using ambulatory wrist actigraphy monitors
(Gaehwiler Electronics, Switzerland) worn on the nondominant hand. Rest-activity data were collected in
30-second epochs. The amount of activity was defined
as the proportion of the epoch spent moving on a scale
from 0 to 255. To control for the activity associated
with capsule administration during the sleep period,
the activity bout associated with each arousal at 1100
and 1400 hours was edited from the raw data. Restactivity data for all groups were then transformed into
alternating periods of arousal and nonarousal, using
an algorithm that determined whether contiguous nonzero epochs of activity were above or below a preset
threshold based on the amount and duration of activity
within a specific activity bout. This algorithm has been
validated against polysomnographic records in a similar subject population and showed average correlations between 0.85 and 0.9 for actigraphic and electroencephalographic (EEG) measures of wakefulness
(26). According to this model, the amount of "energy"
is derived from the product of the Root Mean Square
(RMS) activity level and the duration of the arousal
(or nonarousal) period. In broad terms, the level of
energy in an arousal period reflects the level of wakefulness. Higher levels of energy in an arousal period
reflect an increase in the degree of wakefulness. In the
same manner, the amount of energy in nonarousal periods reflects the depth of sleep. Lower levels of energy
in a nonarousal period reflect a deeper level of sleep,
as evidenced by a greater proportion of stages 3 and 4
relative to stages 1 and 2 (26).
Actigraphy data for each sleep period were extracted
for each subject, transformed and four measures were
determined. The first was a Movement Index (M!),
which indicates the percentage of the sleep period in
which the subject is active, and is highly correlated
with sleep efficiency (for review see reference 27). The
second measure is the total of all energy recorded during the 8-hour sleep period. The third and fourth measures are derived by using the threshold from the scoring algorithm to divide the total amount of energy into
the component amounts of energy in the arousal and
non arousal periods, respectively.
Each of these four measures were analyzed using a
two-factor repeated measures ANOV A. The first factor
was treatment group (bright light vs. melatonin vs.
placebo); the second factor was time (day sleep 1, 2
and 3). Actigraphic data on sleep quality are reported
for the entire group, except for the data for one sleep
Cognitive performance was measured using a computer-based divided attention task that involved simultaneous presentation of two discrimination tasks.
The primary task was a Manikin spatial reasoning task;
the second was a verbal reasoning task based on a
Posner same-different letter presentation task (for review of specific tests see reference 28). Performance
was assessed using four measures considered representative of overall performance. The fist two measures
were the mean response times for the primary and
secondary tasks. The third measure (throughput) corrects the raw response time on the primary task for the
error rate by dividing the mean correct response time
by the percentage of correct responses. This measure
controls for differences in speed and accuracy of response. In addition, the variability in response times
was also determined using the standard deviation of
the raw response times on the secondary task.
The performance data for each subject were averaged
across the shift and analyzed using a two-factor repeated-measures ANOV A. The first factor was group
(bright light, melatonin and placebo); the second factor
was time (night shift 1, 2 and 3).
Sleep, Vol. 18, No.1, 1995
RESULTS
Data from the two placebo control groups were not
significantly different on any ofthe outcome measures.
Consequently, their data were collapsed into a single
placebo group of 16 subjects.
Melatonin
The onset time for the melatonin rhythm prior to
the night-shift transition was calculated by averaging
the onset time from nights one and two. The onset
time on the final day of the study was derived from
the complete 24-hour melatonin profile. Onset times
for each individual were obtained before and after the
transition to night shift and were compared using a
two-factor repeated-measures ANOV A. The first factor was treatment condition (placebo vs. bright light
vs. melatonin), the second was time (pre- and postshift
transition).
There was a significant treatment by time interaction
effect for the onset of the melatonin rhythm [F(2,32)
= 19.8, p < 0.001]. Planned comparisons at the 0.05
level showed that all of the groups showed significant
phase shifts between pre- and postshift conditions.
15
BRIGHT LIGHT VERSUS MELATONIN
-0- Control
-+- Ught
10
1:
Ol
·c
-II- Melatonin
8
"0
E
6
2
,.
Q)
.~
~
['l
4
2
OJ
0
.c
TABLE 1. Mean (standard deviation) core temperature
across each of the three sleep periods for the light, melatonin
and placebo groups. Means for each of the treatment groups
were compared using repeated-measures ANOVA. Planned
comparisons significant at the 0.05 level are indicated
."'
0
F
-2
Group
Sleep period 1
Sleep period 2
Sleep period 3
Placebo
Melatonin
Light
37.79 (0.08)
37.71 (0.16)
37.74 (0.08)
37.78 (0.07)
37.71 (0.09)
37.51 * (0.09)
37.73 (0.09)
37.49* (0.10)
37.46* (0.10)
*p
< 0.05.
Q)
E
-4
Pre-
Post-
FIG. 2. Mean times of the dim-light melatonin onset (DLMO) for
each of the three groups (bright light, melatonin and control) prior
to, and following a 3-day transition to night shift. Pre-shift transition
onsets were the average of the onset obtained on days I and 2 of
the study. Postshift DLMO is taken from the 24-hour period immediately following the last day sleep period. Error bars indicate 1
SEM.
However, there was no difference between the placebo
and melatonin groups, which shifted 4.2 hours (SD =
1.6 hours) and 4.7 hours (SD = 1.2 hours), respectively.
In contrast,the bright-light-treated group showed a significantly greater mean shift of 8.8 hours (SD = 1.5
hours), as seen in Fig. 2.
Core temperature
Table 1 shows the mean core temperature and standard deviations across each of the three sleep periods
for the light, melatonin and placebo groups. Statistical
comparison showed a significant group by time interaction [F(4,64) = 3.03, p < 0.05]. Post hoc comparisons at the 0.05 level indicated that the light group had
a significantly lower temperature than the placebo group
on the second and third sleep periods, whereas the
melatonin group was lower only on the third sleep
period.
To compare differences between groups in the course
of core body temperature for each of the three sleep
periods, data were averaged into 30-minute bins and
analyzed using a repeated-measures ANOVA. Although there were no significant group by time interaction effects for these data, there were some interesting
qualitative differences in the time course of core temperatures between the two treatment groups.
For example, at the start of the third sleep period,
core temperature in the melatonin group was similar
to that in the placebo group but rapidly declined to
that observed in the light-treatment group. Temperature in the melatonin group then stayed relatively constant across the remainder ofthe sleep period. In contrast, core temperature in the light group was lower at
sleep onset, declined to a lower value in the first half
of the sleep period, then increased to approximately
the same value as the melatonin group for the second
half of the sleep period (Fig. 3).
Sleep quality
Overall, the two treatment groups showed marked
improvements in sleep quality relative to the placebo
group. For both treatment groups, sleep was most improved for the second day-sleep period. Across most
of the measures, the light group showed the greatest
improvement, whereas the melatonin group showed
significant but less impressive improvements. These
differences are illustrated in Fig. 4.
Statistical comparison of the MI showed a significant
group by time interaction effect [F(4,64) = 3.3, p <
0.02]. Planned comparisons at the 0.05 level indicated
a nonsignificant reduction in the MI for the first daysleep period for the melatonin and light groups relative
to the placebo group. The mean MI for the placebo
group was 15.5% (SD = 5.7%) compared with 12.3%
(SD = 3.3%) and 12.6% (SD = 7.9%) for the melatonin
and bright-light groups, respectively. For the second
day-sleep period this difference increased. The MI for
the placebo group increased significantly to 24.2% (SD
= 7. 1%), whereas there was a nonsignificant increase
to 14.3% and 13.8% in the melatonin and light groups,
respectively. By the third day-sleep period, the MI for
the placebo group decreased, eliminating most of the
difference between the treatment and placebo groups.
Nevertheless, the light group still showed a significantly lower MI than the placebo group. The mean MI
on day-sleep 3 was 17.6% (SD = 4.7%) for the placebo
group and 13.4% (SD = 4.8%) for the light group. In
contrast, the mean MI for the melatonin group was at
an intermediate value of 15.2% (SD = 7.9%), which
was not significantly different from either the placebo
or light conditions.
Total activity across the sleep period was determined
from the "total energy" measure. There were significant main effects for group [F(2,32) = 6.1, p < 0.01]
and time [F(4,64) = 9.3, p < 0.01]. Post hoc comparisons at the 0.05 level showed that the light and melatonin treatment groups had significantly lower amounts
Sleep, Vol. 18. No.1, 1995
D. DA WSON ET AL.
16
Day 1
o Placebo
• Melatonin
o Light
38
379
2B
26
37.8
<l>
24
~
22
Movement Index
c
~><l>
~ 37.7
o
jij
:;
'" 37.6
~
-0- Control
........ Melatonin
20
18
___ Light
16
14
Q)
'"0,
~
12
375
10
37.4
Sleep penod 1
Sleep period 2
Sleep period 3
37.3
372 81---9--1~0--1~1--1~2--1~3--1~4--1~5--1~6--1"'7-~18
Time of day
1400
1300
1200
~
38
DAY 2
o Placebo
37.9
• Melatonin
o light
37.8
§
1100
C'
1000
I
Total Energy Count
-0- Control
........ Melatonin
___ Lighl
900
800
700
600
~
377
500~~~==~~~==~~~~~~~
Sleep period 1
Sleep period 2
Sleep period 3
0;
<.)
•
376
~
o
37.5
37.4
1200
37.3
1100
37.21--~---~---~-~--.::--~--:.,..--:,.
8
9
10
11
12
13
14
15
16
17
Energy in Arousal periods
IS
--0- Control
.....- Melatonin
-to- Lighl
Time of day
500
38
Day 3
400~~====~~~~==~~~====~
Sleep period 3
o Placebo
37.9
• Melatonin
o light
37.8
260
~ 37.7
.u;
0;
240
~ 376
220
'"~
Total Energy in non·arousals
g' 375
-0- Placebo
o
-+-
Melatonin
___ light
37.4
37.3
120
9
10
11
12
13
14
Time of day
15
16
17
18
FIG. 3. Core temp in degrees Celsius across the sleep period for
each of the three groups (bright light, melatonin and control) for
each of the 3-day sleep periods. Core temperature was binned into
3D-minute intervals then averaged across subjects. Error bars indicate 1 SEM.
Sleep, Vol. 18, No.1, 1995
100~======~~~=====?~~====~
Sleep period 1
Sleep period 2
Sleep period 3
FIG. 4. Arousal data for the three groups (bright light, melatonin
and control) across each of the 3-day sleep periods. The top graph
indicates the movement index (M!), which measures the proportion
of the sleep period spent aroused. The second graph indicates the
total amount of energy expended across the sleep period. Total energy
was then divided into the amount of energy in arousal and nonarousal periods, respectively, according to an algorithm defined previously. Thus, the third graph indicates total of energy in arousal
periods and the fourth indicates the amount of energy in nonarousal
periods. Energy is given in arbitrary units derived from the product
of the duration of the period and the mean RMS amplitude of activity
during an arousal or nonarousal period. Error bars indicate 1 SEM.
17
BRIGHT LIGHT VERSUS MELATONIN
1800
800
1600
750
700
oQ)
~
posner S-D AT
650
-0- Placebo
600
-.- Melatonin
1400
Manikin AT
1200
-0- Placebo
-+- Melalonin
___ Ughl
0
'"
E
(f)
___ Ughl
550
1000
500
800
450
400
600
-'-t::==::::=::==::r-===:=.=:===r--=:::::;;::=.:=:;:=-
Shift 1
2000
450
1600
400
Manikin Throughput
0
-0- Placebo
Q)
E
Shift 3
500
1800
(f)
Shift 2
1400
-.- Melatonin
___ Ughl
Posner Variability
0
'"
350
-0- Placebo
(f)
E
-+- Melatonin
___ Ught
300
1200
250
1000
200
800
Shift I
Shift 2
Shift 3
150
Shift 1
Shift 2
Shift 3
FIG. 5. Mean cognitive psychomotor performance for each of the groups across each of the three shifts for each of four measures. Top
left indicates response latency for the verbal reasoning task. Top right indicates response latencies for the spatial reasoning task. Bottom
left indicates throughput for the spatial reasoning task. Bottom right indicates the variability (SD of mean response latency) in response
latencies for the verbal reasoning task. Response latencies are all given in milliseconds, throughput is in milliseconds corrected for the
proportion correct. Error bars indicate 1 SEM.
of energy compared to the placebo group. Furthermore,
there was no difference in total energy between the two
treatment groups. Planned comparison indicated that
the placebo group showed a consistent increase in energy across each of the sleep periods, whereas the two
treatment groups showed no increase. These differences are illustrated in Fig. 4.
The total amount of energy in the sleep period was
then divided into arousal and nonarousal periods, which
were considered independently. The light group had
consistently less energy in both arousal and nonarousal
periods than the placebo group, and the melatonin
group showed intermediate reductions in energy levels.
These data are illustrated in Fig. 4.
Statistical comparison of the mean energy in all
arousals showed a significant main effect for group
[F(2,32) = 3.4, p < 0.05]. Fisher PLSD post hoc comparisons at the 0.05 level indicated that the mean energy in all arousals was significantly lower in the light
group (mean 588.7, SD = 247.1) than in the melatonin
group (mean = 745.7, SD = 313.1), which was, in tum,
significantly lower than for the placebo group (mean
= 900.1, SD = 362.2).
A comparison of the mean energy in ali nonarousal
periods showed a similar result. There was a significant
main effect for group [F(2,32) = 3.8, p < 0.05]. Post
hoc comparisons at the 0.05 level showed that the light
treatment group had the lowest mean activity levels
during the periods defined as nonarousal (mean = 142.5,
SD = 64.5). The placebo group showed the highest
mean values (mean = 212.0, SD = 96.9) and, as with
mean energy in the arousal periods, the melatonin group
showed intermediate values, which were not significantly different to either group (mean = 171.5, SD =
73.0).
Cognitive performance
Compared to the placebo group, the light treatment
group consistently showed an improvement in cognitive performance across the three night shifts. These
differences were, however, relatively small and not always significant across all variables. In contrast, the
melatonin treatment group typically performed at the
same level as the control group. Figure 5 displays the
results for each of the four individual performance
measures.
Mean correct response times for the verbal reasoning
task indicated that there was a significant main effect
for time [F(2,32) = 85.6, p < 0.001] and a trend for
Sleep, Vol. 18, No. I, 1995
D. DA WSON ET AL.
18
a group effect [F(2,32) = 2.52, p < 0.1]. Planned comparisons at the 0.05 level showed that mean correct
response times in all groups decreased significantly
across all three night shifts. Post hoc comparisons between each of the treatment groups and the placebo
group indicated that the light group responded significantly faster than the placebo and melatonin groups
across all three night shifts. There were no differences
in mean correct response times between the melatonin
and placebo groups on any of the shifts.
Mean correct response time for the spatial reasoning
task showed the light-group subjects to be consistently
faster in their response time than the melatonin or
placebo groups. There was a significant group by time
interaction effect [F(2,32) = 4.0, p < 0.01]. Planned
comparisons at the 0.05 level showed no significant
differences between any ofthe groups on the first night
shift. On the second and third shifts, the light group
showed significantly faster mean correct response times
than the placebo group, with the melatonin treatment
group again showing nonsignificant intermediate improvements on the second and third night shifts.
When response times were corrected for error rates
to give a throughput measure, there was a significant
main effect for time [F(2,32) = 66.2, p < 0.001]. This
indicated that the significant improvement in response
times in the two treatment groups was attenuated by
a speed-accuracy trade-off. When the variability in the
response rate was examined, there was a significant
main effect for time [F(2,32) = 16.7, p < 0.001] and
a trend for a group effect [F(2,32) = 2.56, p < 0.1].
Figure 5 indicates that the melatonin and placebo
groups showed a similar downward trend in variability
across the study. The light group showed a similar
trend, but response latencies were consistently lower
across the study.
DISCUSSION
These data indicate that three pulses of bright light
administered between midnight and 0400 hours produce significantly greater phase shifts in the circadian
system than daytime melatonin administration or dim
red light. On average, the light-treatment group phase
shifted the onset of their melatonin rhythm by 8-9
hours. In contrast, the placebo and melatonin groups
shifted only about 4 hours. This study, as well as others
(3-5,29), clearly demonstrates that exposure to appropriately timed bright light can improve circadian adaptation. By comparison, a similar 3-day protocol using only a single pulse of bright light (4) produced a
mean phase shift of approximately 4-5 hours. Thus,
the use of two additional exposures nearly doubled the
size of the phase shift.
It is worth noting that in this study, the placebo
Sleep, Vol. 18, No.1, 1995
group showed a significant phase shift relative to the
baseline position for their melatonin rhythm. This is
in contrast to previous studies (e.g. 3,30) that have
emphasized the lack of circadian adaptation in night
workers. In the two previously cited studies, subjects
were not restricted to a laboratory setting during the
study and commuted home following the end of the
shift. As a consequence, the failure to adapt may reflect
whether subjects are exposed to bright light prior to
going to bed. In fact, a recent report has suggested that
attenuating early morning exposure to light with dark
glasses or goggles alone may significantly increase circadian adaptation independent of bright-light exposure
during the shift (31,32).
Although there was a significant increase in the extent
of the phase shift for the light treatment group, there
was no difference in the phase shift between the melatonin and placebo groups. This result was unexpected
because subjects received the first dose of melatonin at
a time that has been predicted to produce delays in the
circadian system (12). However, the administration
protocol employed in this study is likely to have produced plasma melatonin levels at or above normal physiologicallevels from 0900 hours to 1700 hours, possibly
producing a concurrent phase advance. Thus, it is possible that the lack of any significant phase shift may
reflect our specific melatonin protocol, producing simultaneous phase advances and phase delays that result
in a relatively small net shift. A single administration
at sleep onset may produce more substantial delays (12).
However, the relatively short half-life of melatonin
would result in an expected hypothermic effect, particularly in the second half of the sleep period. Given the
relatively modest phase-shifting properties of melatonin
over 3 days of administration (12), the increase in the
phase shift may not compensate for the loss of the hypothermic effect. However, this remains an empirical
question.
Despite significant differences in the degree of the
circadian phase shift, when compared with the placebo
group, both treatments produced significant reductions
in core temperature by the third sleep period. Although
the melatonin administration protocol employed in
this study had little effect on the phase of the endogenous melatonin rhythm, the hypothermic effect produced a similar outcome. By dropping core temperature through a noncircadian mechanism, melatonin
produced a similar outcome to phase shifting; that is,
it permitted subjects to reduce core temperature across
the sleep period.
When averaged across the whole sleep period, the
light-treatment group had significantly lower mean core
temperature by the second sleep period and the melatonin group by the third sleep period. However, these
differences were not consistent over the entire sleep
t
BRIGHT LIGHT VERSUS MELATONIN
period. As Fig. 3 indicates, the melatonin group had
a slightly lower temperature than the light group for
the first half of the first sleep period but not during the
second half. By contrast, core temperature in the light
group was similar to the placebo group across the entire
first sleep period, yet consistently lower across the second and third sleep periods.
By the third sleep period, core temperature for the
melatonin and light groups was consistently lower
across the entire sleep period except for the first 30
minutes of the sleep period. Although core temperature
in the melatonin and placebo groups was typically the
same at the start of the sleep period, the core temperature in these groups dropped rapidly to the level observed in the light-treatment group. This rapid decline
may reflect the onset of the hypothermic action as
plasma melatonin levels increase to, or beyond, physiological levels.
The drop in core temperature across the sleep period
for both treatment groups was associated with significant improvements in the actigraphic measures of sleep
quality. Both treatment groups showed significant reductions in the proportion of the sleep period spent
aroused, particularly for the second sleep period. Because sleep quality in the second sleep period following
a night-shift transition is often worse than the first, the
lack of a significant difference in the MI between the
treatment and placebo groups for the first sleep period
may reflect increased sleep pressure due to the extended period of wakefulness (24 hours) prior to the first
sleep period. The reduction in the difference in the MI
between the second and third sleep period for the placebo and treatment groups was primarily due to improvement by the placebo group. This suggests that by
the third day of the night-shift transition, the placebo
group is likely to have shifted phase sufficiently for
sleep quality and duration to start improving.
Whereas the two treatments produce similar improvements in the duration of the nonarousal period,
a more detailed analysis of the actigraphic data indicated that the light group, although aroused for a similar amount of time to the melatonin group, was significantly less active during these arousals. This suggests
that sleep quality was better for the light group than
for the melatonin group. Across each of the three sleep
quality indices, the light group showed significantly
lower levels of energy in both arousal and nonarousal
periods than the placebo groups. In contrast, intermediate decreases in energy levels were observed in
the melatonin group. Typically, the melatonin group
showed 50-60% of the decrease reported by the lighttreatment group.
Whereas the melatonin and light treatment groups
both reported increases in the duration and quality of
sleep, only the light group showed consistently higher
19
performance levels across the night. Similarly, several
of the measures indicated only trends to a significant
difference between the groups. However, this may reflect the relatively low number of subjects in the lighttreatment group rather than a lack of effect. Many of
the differences were in the predicted direction and paralleled differences in circadian adaptation or core temperature across the study.
For example, compared with the placebo group,
complex spatial reasoning and simple verbal reasoning
were improved in the light-treatment group but not in
the melatonin group. Despite having lower activity
levels across the sleep period than the placebo group,
the melatonin group did not show a concomitant improvement in performance. This is likely to reflect differences between the two groups while on shift. To
phase shift the circadian system, subjects in the light
group were exposed to bright light while on shift. This
exposure is likely to have enhanced cognitive psychomotor performance because nocturnal exposure to
lighting greater than 1,000 lux has been shown to enhance performance independent ofthe circadian effects
(33,34). In addition, exposure at 4-700 lux is likely to
have suppressed melatonin levels (35). The melatonin
group did not shift their melatonin rhythm to the same
extent as the light-treatment group; therefore, their circulating melatonin levels at night were likely to be
similar to those in the control group. Since melatonin
has been shown to have hypothermic effects (19-22)
and to reduce simple response times (36), the suppression of melatonin in the light group across the night
shift may have improved on-shift performance through
tonic as well as circadian effects.
Taken together, these results clearly indicate that
appropriately timed exposure to bright light can significantly improve circadian adaptation to simulated
night shift. This increase in the rate of adaptation is
associated with significant improvements in the duration and quality of actigraphically measured sleep
and some measures of cognitive psychomotor performance. In contrast, daytime administration of melatonin did not increase the rate of circadian adaptation
but did produce intermediate improvements in sleep
and some performance measures, although this is likely
to reflect hypothermic rather than circadian effects.
On the basis of these data, bright light would appear
to be a better short-term treatment for improving adaptation to night shift. It may not, however, always be
the best approach. Although subjects achieve significantly greater phase shifts during the night shift, they
will also require greater phase shifts to readapt to the
day shift. In the long term this may be problematic
because it is not yet known whether the long-term
morbidity associated with shift work (37-39) reflects
the amount of time spent in a de synchronized state or
Sleep. Vol. 18. No.1. 1995
.~
20
D. DA WSON ET AL.
the overall phase disturbance to the circadian system.
If it is the degree to which the circadian system is
disturbed that produces long-term negative effects, interventions that result in greater total shifts in the circadian system (i.e. bright artificial light) may in the
long run prove to be counterproductive. There was
some limited, albeit anecdotal, support for this in the
subjective reports of the readaptation process. The subjects receiving bright light treatment reported greater
difficulty readjusting to day sleep than those receiving
melatonin. In addition, many of the bright-light group
reported paradoxical feelings while on night shift. Although they performed better and slept better, they
often reported feeling "pushed" and fatigued. Although
we have no clear objective basis for these interpretations, we believe it is important to emphasize the light
group's relatively frequent subjective reports offeeling
"stressed" or "anxious". Interestingly, these feelings
were similar to those often reported in jet-lag studies.
Perhaps many of the negative mood changes associated
with circadian disruption reflect the degree of circadian
stress imposed by the adaptation in phase rather than
the inertia of the circadian system.
Until now, the inertia of the circadian system in shift
workers has always been viewed as maladaptive. If,
however, it is the degree of pacemaker instability induced by shift changes that is "toxic", the inertia of
the circadian system may, in fact, be protective. In this
case, treatment strategies that produce large net shifts
in the circadian system (e.g. nocturnal bright-light exposure) may be beneficial in the short term yet "chronotoxic" when used for extended periods.
An alternative approach, particularly for rapidly rotating shift schedules might be to temporarily disconnect the "clock" regulating core temperature and sleep.
By simply reducing core temperature during the daytime sleep period, it may be possible to reduce arousals
related to the diurnal increase in core temperature produced by the circadian system. In this study, exogenous
melatonin had a significant thermodepressive effect associated with improved sleep. If melatonin were used
to inhibit, or mask, the pacemaker-mediated rise in
core temperature that is thought to truncate sleep, it
might be used to facilitate sleep without requiring the
large phase changes induced by bright-light exposures.
Such a strategy would permit shift workers to "override" the circadian system for short periods and avoid
the potential toxicity due to overzealous manipulations
of the circadian pacemaker. This would be particularly
useful in shift schedules that rotate rapidly or where
timed exposure to bright light is impractical.
Acknowledgement: This research was supported by the
Mining and Quarrying Industry Occupational Health and
Safety Committee, Adelaide, Australia.
Sleep. Vol. 18, No.1, 1995
REFERENCES
I. Moore-Ede MC, Sulzman FM, Fuller CA. The clocks that time
us. Cambridge, MA: Harvard University Press, 1982.
2. U.S. Congress, Office of Technology Assessment. Biological
rhythms: implicationsjor the worker. OTA-BA-463. Washington
DC: Government Printing Office, 1991.
3. Czeisler CA, Johnson MP, Duffy JF, Brown EN, Ronda IS,
Kronauer RE. Exposure to bright light and darkness to treat
physiologic maladaptation to night work. New Engl J Med 1990;
322(18):1253-9.
4. Dawson D, Campbell SS. Timed exposure to bright light improves sleep and alertness during simulated night shifts. Sleep
1991;14:511-6.
5. Eastman Cl. High-intensity light for circadian adaptation to a
12-h shift of the sleep schedule. Am J PhysioI1992;263:R42836.
6. Czeisler CA, Allan IS, Strogatz SH, et al. Bright light resets the
human circadian pacemaker independent of the timing of the
sleeIr-wake cycle. Science 1986;233(4764):667-71.
7. CzeislerCA, Kronauer RE, Allan JS, et a!. Bright light induction
of strong (type 0) resetting of the human circadian pacemaker.
Science 1989;244(4910): 1328-33.
8. Dawson D, Lack LC, Morris M. Phase resetting of the human
circadian pacemaker with use of a single pulse of bright light.
ChronobiolInt 1993;10(2):94-102.
9. Dijk DJ, Beersma DG, Daan S, Lewy AJ. Bright morning light
advances the human circadian system without affecting NREM
sleep homeostasis. Am J PhysioI1989;256:RI06-11.
10. Drennan M, Kripke DF, Gillin Je. Bright light can delay human
temperature rhythm independent of sleep. Am J Physiol1989;
257:RI36-41.
II. Czeisler CA, Weitzman ED, Moore-Ede MC, Zimmerman JC,
Knauer RS. Human sleep: its duration and organization depend
on its circadian phase. Science 1980;210(4475):1264-7.
12. Lewy AJ, Ahmed S, Latham Jackson 1M, Sack RL. Melatonin
shifts human circadian rhythms according to a phase-response
curve. Chronobiol Int 1992;9(5):380-92.
13. Redman J, Armstrong S, Ng KT. Free-running activity rhythms
in the rat: entrainment by melatonin. Science 1983;219:108991.
14. ArendtJ, Bojkowski C, Folkards, eta!. Some effects of melatonin
and the control of its secretion in humans. Ciba Found Symp
1985; 117:266-83.
15. ArendtJ, Borbely AA, FraneyC, WrightJ. The effects of chronic,
small doses of melatonin given in the late afternoon on fatigue
in man: a preliminary study. Neurosci Lett 1984;45:317-21.
16. Cramer H, Rudolph J, Consbruch U. On the effects of melatonin
on sleep and behaviour in man. In: Costa E, Gessa GL, Sandler
M, eds. Serotonin- new vistas: biochemistry and behavioural and
clinical studies. Advances in biochemical psychopharmacology,
vol. II. New York: Raven Press, 1974:187-91.
17. Lieberman HR. Behaviour, sleep and melatonin. J Neural
Transm 1986;Suppl 21 :233-41.
18. LiebermanHR, WaldhauserF, Garfield G, Lynch HJ, Wurtman
RJ. Effects of melatonin on human mood and performance.
Brain Res 1984;323:201-7.
19. Arutyunyan G, Mashdovskii M, Roshchina L. Pharmacological
properties of melatonin. Fed Proc 1964;23:T1330-2.
20. Cagnacci A, Elliot JA, Yen SSe. Melatonin: a major regulator
of the circadian rhythm of core temperature in humans. J Clin
Endocrinol Metab 1992;75:447-52.
21. Carman JS, Post RM, Buswell R, Goodwin FK. Negative effects
of melatonin on depression. Am J Psychiatry 1976;133:1181-6.
22. Dawson D, Singh P, Encel N, Govas K, Matthews C. Exogenous
melatonin decreases core body temperature. In: Topliss Dl, ed.
The endocrine society ojAustralia proceedings. 35th annual scientific meeting, 1992:88.
23. Dawson D, Encel N. Melatonin and sleep in humans. J Pin Res
1993;15:1-12.
24. Lewy Al, Sack RL. The dim light melatonin onset as a marker
for circadian phase position. ChronobiolInt 1989;6:93-102.
BRIGHT LIGHT VERSUS MELATONIN
25. Earle CR, D'Occhio MJ, Kennaway DJ, Seamark RF. Serum
melatonin profiles and endocrine responses of ewes exposed to
a pulse oflight late in the dark phase. Endocrinology 1985; 117:
226-30.
26. Wallace G, Dawson D, Filsell S. Wrist actigraphy as a predictor
of sleep-wake behaviour. Aus-Sleep '92, Proceedings of the Fifth
Australian Sleep Research Meeting, Cairns, 1992.
27. Tryon, W. Activity measurement in psychology and medicine.
New York: Plenum Press, 1991.
28. Thorne D, Genser S, Singer H, Hegge F. The Walter Reed performance assessment battery. Neurobeh Toxieol Teratal 1985;
7:415-8.
29. Eastman CI, Miescke K. Entrainment of circadian rhythms with
26-h bright light and sleep-wake schedule. Am J Physiol 1990;
259:RI189-97.
30. van Loon JH. Diurnal body temperature curves in shift workers.
Ergonomics 1963;6:267-73.
31. Eastman CI. Dark goggles during daylight and high-intensity
light during night shifts for circadian adaptation to simulated
shiftwork. Sleep Res 1993;23:402.
32. Wisbey J, Budensiek D. Light therapy in a sleep lab technologist:
a case study. Sleep Res 1993;23:292.
21
33. Badia P, Myers B, Boecker M, Culpepper J, Harsh J. Bright·
light effects on body temperature, alertness, EEG and behaviour.
Physiol Behav 1991 ;50(3):583-8.
34. Campbell SS, Dawson D. Enhancement of night-time alertness
and performance with bright ambient light. Physiol Behav 1990;
48:317-20.
35. Lewy AJ, Wehr TA, Goodwin FK, Newsome DA, Markey SP.
Light suppresses melatonin secretion in humans. Science 1980;
210:1267-9.
36. Wynn VT, Arendt J. Effect of melatonin on the human electrocardiogram and simple reaction time responses. J Pineal Res
1988;5:427-35.
37. Angerspach D, Knauth P, Loskant H. A retrospective cohort
study comparing complaints and diseases in day and shift workers. Int Arch Occup Environ Health 1980;45: 127-40.
38. Aanonsen A. Shift work and health. Oslo, Norway: Universitetsforlaget, 1964.
39. Rutenfranz J, Colquhoun WP, Knauth P, Garter J. Biomedical
and psychological aspects of shift work. Scan J Work Environ
Health 1977;3:165-82.
Sleep, Vol. 18, No.1, 1995