Sleep-wake differences in heart rate variability during a 105-day simulated mission to
Daniel E. Vigo, M.D., Ph.D. (1,2), Barbara Ogrinz, M.A. (2), Li Wan, M.D., Ph.D. (2), Evgeny Bersenev, Ph.D.
(3), Francis Tuerlinckx Ph.D. (2), Omer Van den Bergh, Ph.D. (2) and André E. Aubert, M.D., Ph.D. (4)
(1) Departamento de Docencia e Investigación, Facultad de Ciencias Médicas, Universidad Católica
Argentina and Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Buenos Aires,
Argentina; (2) Department of Psychology, University of Leuven, Leuven, Belgium; (3) State Science Center of
Russian Federation - Institute of Biomedical Problems of Russian Academia of Science, Moscow, Russia; (4)
Laboratory of Experimental Cardiology and Interdisciplinary Centre for Space Studies, University of Leuven,
Dr. André E. Aubert
Laboratory of Experimental Cardiology and Interdisciplinary Centre for Space Studies
Katholieke Universiteit Leuven
Herestraat 49; B-3000 LEUVEN, Belgium
Tel: 32 16 345841, Fax: 32 16 345844
Short title: Autonomic activity in Mars105
Word count for abstract: 250, word count for main text: 3139, number of references: 27, number of tables: 2,
number of figures: 2.
Introduction: In prolonged space flights the effect of long-term confinement on the
autonomic regulation of the heart is difficult to separate from the effect of prolonged
exposure to microgravity or other space-related stressors. Our objective was to investigate
whether the sleep-wake variations in the autonomic control of the heart are specifically
altered by long-term confinement during the 105-day pilot study of the earth-based Mars500
project. Methods: Twenty-four-hour EKG records were obtained before (pre), during (T1:
30, T2: 70 and T3: 100 days), and after (post) confinement in the six crew members that
participated in the mission. Sleep and wake periods were determined by fitting a square
wave to the data. Autonomic activity was evaluated through time and frequency domain
indexes of heart rate variability (HRV) analysis in wake and sleep periods. Results: During
confinement, wake HRV showed decreased mean heart rate and increased amplitude at all
frequency levels, particularly in the very low (pre: 13.3 ± 0.2; T1: 13.9 ± 0.3; T2: 13.9 ± 0.2;
T3: 13.9 ± 0.2; post: 13.2 ± 0.2), and high (pre: 7.6 ± 0.4; T1: 8.3 ± 0.5; T2: 8.2 ± 0.4; T3:
8.1 ± 0.4; post: 7.6 ± 0.3) frequency components (values expressed as mean ± SE of
wavelet power coefficients). Sleep HRV remained constant, while sleep-wake high
frequency HRV differences diminished. Discussion: The observed autonomic changes
during confinement reflect an increase in parasympathetic activity during wake periods.
Several factors could account for this observation, including reduced daylight exposure
related to the confinement situation.
Autonomic nervous system, confinement, Mars500, space physiology
Several environmental factors related to long-term space flight, like confinement,
varying light exposure, noise or temperature changes, absence of earth-based Zeitgeber,
are likely to play a role in sleep – wake cycle alterations (14). Some evidence has been
found about the effect of confinement by itself on circadian rhythms. A 28-day isolation
experiment revealed no major sleep disturbances besides an increase in self-rated
tiredness (22). In contrast, during a 7-day confinement period, catecholamines and sleep
motor activity exhibited significant increases (13). Moreover, adaptation to changed
physical and social environments during isolation periods of 135 days was associated with
specific changes in sleep architecture (25).
Although autonomic nervous system (ANS) activity may play a key role in
performance during space missions (3), little is known about the specific impact of
confinement on the circadian rhythm of ANS activity. This rhythm is characterized by a
sympathetic predominance during the wake periods that allows an active engagement with
the external environment with increased utilization of energy, and a parasympathetic
predominance during the night related with a disengagement from external environment for
recovery (18). A decrease in parasympathetic activity was reported in space (1) and ground
(23) based bed-rest experiments with long periods of confinement. In an experiment
conducted in order to characterize neurovegetative activity in a ground based unit that
simulated the living conditions of a space station except microgravity, no major differences
were disclosed before, during and after the isolation period of 60 days (16). Still, none of
these studies evaluated the circadian rhythm of autonomic activity (1;16;23).
Hence, we sought to investigate the circadian profile of heart rate variability during a
105-day confinement period in the context of the Mars500 pilot study (Mars105). The aim of
the Mars500 project is to gather data, knowledge and experience about the psychological
and physiological effects of living in an earth-based enclosed environment during the 520
days as would be required for a real mission to Mars. This allows to separate the effect of
long term confinement from the effect of long term exposure to microgravity. We
hypothesize that sleep-wake variations of the autonomic control of the heart are altered by
long term confinement.
Six healthy non-smoking male subjects (mean ± SD: age 33 ± 6 years; height 181 ±
5 cm; weight 82 ± 12 kg; BMI 25 ± 3 kg/m2) were selected to participate in a 105-day
confinement pilot-study before the Mars500 project.
The Mars 500 project, organized by the European Space Agency (ESA) and the
Institute for Biomedical Problems (IBMP) at Moscow, is designed to simulate a mission to
Mars in duration, composition of the crew, activities, work load and communication facilities.
The protocol of the study reported herein was approved in advance by the Ethics
Committee of the University Hospital Gasthuisberg of Leuven, Belgium and the ESA
Medical Board, which complied with all guidelines stated in the Declaration of Helsinki. All
participants gave informed consent to participate in the study.
Subjects were confined in the isolation facility at IBMP in Moscow from the 31st of
March 2009 to the 14th of July 2009. The lay-out of the isolation facility comprises 4
hermetically sealed interconnected habitat modules with artificial lighting conditions (50 –
300 lux). The total volume of the habitat modules is 550 m3. Ambient temperature was
maintained constant at 24 °C, with a relative humidity of 35-45%. Subjects were involved in
different scientific protocols to assess the psychological and physiological effects of
isolation and confinement. Their schedules were organized in order to maintain 8-hour
periods of work, leisure and sleep. Crewmembers operated on night-shifts for one week
each, in rotation.
Twenty-four hour Holter signals were obtained at five time points: in one day
between 17 to 20 days before confinement (Pre); in one day between the 38th to the 40th
day of confinement (T1); in one day between the 73th to the 76th day of confinement (T2);
in one day between the 98th to the 100th day of confinement (T3); and in one day between
11 to 13 days after the end of confinement (Post). Data collection was performed
regardless of day or night shift of the subject.
Signal recording: Electrocardiogram signal was recorded using a digital Holter
device. Ventricular depolarizations (R waves) were detected through the device software.
The time elapsed between R waves (RR intervals) was then computed. Heart rate
variability (HRV) indexes were computed in 1-hour segments. Premature and lost beats
were identified by an automated filter and replaced by RR intervals resulting from linear
Time domain: Quantitative time series analysis was performed on heart rate by
evaluating measures of variation over time. Among these, RRm (mean duration of RR
intervals in ms) quantifies the mean heart rate, SDNN (standard deviation of RR intervals in
ms) represents a coarse quantification of overall variability, and RMSSD (square root of the
mean squared differences of successive normal RR) measures short-term heart rate
Frequency domain: These measurements provide an evaluation of the power of the
contributing frequencies underlying HRV. Its high-frequency (HF) component (0.15-0.4 Hz)
is related to respiratory sinus arrhythmia and mediated by parasympathetic activity,
whereas the low-frequency (LF) component (0.04-0.15 Hz) is related to baroreflex control
and depends upon sympathetic and parasympathetic mechanisms. A very low frequency
(VLF) component (<0.04 Hz) of an uncertain origin is also found and has been attributed to
thermoregulatory fluctuations in vasomotor tone as well as to humoral factors such as the
renin-angiotensin system, with dependence on the presence of parasympathetic outflow
(Figure 1) (20;21).
[Fig. 1 here]
To analyze the frequency components of HRV, the Discrete Wavelet Transform
(DWT) was chosen rather than the traditional Fast Fourier Transform (FFT) because it is
not affected by discontinuities or non-stationarities (2). Before applying the DWT, the linear
trend and the mean value were subtracted from the signal. In addition, it was evenly
sampled with a frequency of 2.4 Hz by means of a spline interpolation algorithm and zero
padded to the next higher power of two (2). A six-level wavelet decomposition was
employed to analyze the signal, using a Daubechies 4 wavelet function. Using this
decomposition, wavelet levels A6 and D1-D6 represent the total power (TP, 0–0.6 Hz),
wavelet levels A6 and D6 approximately correspond to the very low frequency band (VLF,
0-0.0375 Hz), wavelet levels D4-D5 to the low frequency band (LF, 0.0375-0.15 Hz), and
wavelet levels D2-D3 to the high frequency band (HF, 0.15-0.6 Hz). In DWT, the square of
the standard deviation of wavelet coefficients at each level is concordant with the spectral
power of that level (2). Reported values are expressed as the natural logarithm of TP, HF,
LF and VLF; normalized units of LF (LF/(TP–VLF)X100) and HF (HF/(TP–VLF)X100); and
the ratio between LF and HF. The use of normalized units minimizes the effect of the
changes in total power on the values of LF and HF, and emphasizes the balanced behavior
of the two branches of the autonomic nervous system (20).
Subject's reports of waking and sleeping times and actigraphy records were not
available for the pilot study. In general, visual inspection of the individual records showed a
typical fall of heart rate during the night, with abrupt transitions between periods. These
observations suggest a square wave model (two alternating contiguous periods of low and
high heart rate) of the 24-hour heart rate record. These variations are similar to those seen
in 24-hour beat-to-beat blood pressure recordings, where square wave modeling accounted
for a larger fraction of circadian variance than modeling based on visual inspection, cosinor
method or fixed clock time (11).
Briefly, 20-minute consecutive averages of RR-intervals were calculated. Square
waves were constructed using all the possible different combinations of the low- and high-
heart rate periods length. Both the averaged signal and each square wave were
standardized to a mean of zero and a standard deviation of 1.0. Cross correlation values of
the standardized RR-interval signal with all possible different standardized square waves
were determined. The best fitting square wave was identified by the highest cross
correlation value. This square wave was used to segment the original RR-interval record in
a high (wake) and a low (sleep) heart-rate period. The transience time from the high- to the
low- heart rate period was identified as tdown while transience time from the low- to the high-
heart rate period was identified as tup (Figure 2) (11). The fit resulting from this model is
optimal with respect to the square error. The square of the highest cross correlation value
expresses the fraction of total variation (FTV) of the 24-hour blood pressure profile
accounted by the model (11).
[Fig. 2 here]
In four of the thirty records a low heart rate period was identified by the model within
the daytime (before 20:00). Visual inspection of these records revealed two different low-
heart rate periods better described by a biphasic square wave model. Therefore, the
original record was fragmented in a daytime record (before 20:00) and a nighttime record
(after 20:00). The model was applied to these separate fragments in order to finally
construct a biphasic square wave with two different low (sleep) heart rate periods.
Hourly HRV was averaged along wake and night-time sleep periods. HRV
differences between night-sleep and wake averages were also calculated. Each hour was
assigned to a wake or to a sleep period according to the transience times previously
defined. At least 55 minutes of any hour should fall within a specific period to be assigned
to it; otherwise it was marked as a transition hour and excluded from the wake or sleep
period average. Daytime sleep periods were marked as naps and excluded from the wake
Sleep-wake data and HRV indexes were expressed as mean ± standard error.
Normality was assessed by means of a Kolmogorov-Smirnov test. A natural logarithm
transform was used where needed.
Initially, in order to assess the effect of confinement by itself, sleep-wake data, wake
HRV and sleep HRV indexes were averaged along non-confinement days (T1 and T5) and
confinement days (T2, T3, T4). Differences between both conditions were evaluated
through a paired-samples T-test.
Then, differences between measurements days (T1, T2, T3, T4 and T5) were
assessed by means of a repeated measures ANOVA test, followed by a Tukey HSD post-
hoc test. A Mauchly’s sphericity test was conducted in order to use a univariate approach
for ANOVA analysis; when sphericity could not be assumed a multivariate approach was
The mean tdown and tup time points varied non-significantly between midnight and
02:00 AM and between 06:00 AM and 09:00 AM, respectively. These time points
determined sleep periods with a mean duration of six to eight hours. Only subject # 2 wore
the Holter while he was awake on nightshift after 24:00, showing a short night-time sleep
period three hours before midnight, and no day-time sleep period. The fraction of the
variance explained by the model varied non-significantly during isolation between 55% and
75% (not shown).
Confinement was associated with a diminished mean heart rate and an augmented
global HRV during the day. While the increase in HRV was verified in all frequency
components, the comparison between components revealed a relative decrease of LF.
These changes seem to be more evident in the second month of isolation as revealed by
the significant decrease in LF/HF in T2 (Table I).
[Table I here]
During sleep, RR interval duration showed an increase during confinement (pre:
1088 ± 62 ms; T1: 1169 ± 55 ms; T2: 1211 ± 60 ms; T3: 1176 ± 72 ms; post: 1112 ± 58 ms;
F (4, 20) = 3.45; p < 0.05; T2 different from pre and post). However, no significant
differences were found along measurements days in sleep-HRV (not shown).
When comparing wake- and sleep-HRV, it was observed that confinement was
associated with a reduction of HRV sleep-wake differences in SDNN, TP, VLF, LF, HF and
LF/HF. In addition, for TP and VLF, the sign of the mean difference changed (Table II). The
analysis of contrasts between measurement days showed a significant decrease in
RMSSD, HF and LF/HF sleep-wake differences. Post-hoc pairwise comparisons between
measurements days only revealed a significant reduction of HF-HRV sleep-wake
difference, where T1 < post, T2 < pre and T2 < post (Table II).
[Table II here]
The main result of the present study is that during confinement, wake HRV showed
decreased mean heart rate and increased amplitude at all frequency levels, with a
decrease in the normalized units of the low-frequency HRV component. Sleep HRV
remained constant. In addition, confinement was associated with a decrease of VLF-HRV,
LF-HRV, HF-HRV and LF/HF sleep-wake differences and with the appearance of negative
VLF-HRV sleep-wake differences. These changes seem to be more pronounced for HF-
HRV and in the middle of the confinement period. No significant differences were found in
the length or phase of the sleep-wake periods.
Heart rate oscillations at all frequency levels (VLF, LF and HF) reflect
parasympathetic influences (20;21), while LF fluctuations are also tightly coupled with
synchronous oscillations of efferent sympathetic nervous activity (21). Thus, the increase in
HRV at all frequency bands, with a relative decrease in the LF component along
confinement, can be explained by an augmented parasympathetic activity with a loss of
sympathetic predominance during wake periods. The increased vagal predominance during
the day can also account for the vanishing of the sleep-wake differences of SDNN, LF and
HF (usually positive) and LF/HF (usually negative), as well as for the appearance of
negative sleep wake differences of TP and VLF (usually positive) (10).
These results are in line with observations reported in space analogue environments
like Antarctica. In a 40-day stay in the Italian Antarctic Station of Terra Nova Bay, a relative
significant decrease of LF was found, which was interpreted as a reduced sympathetic
activity. This was associated with a significant reduction of the anterior pituitary and adrenal
hormonal levels of the pituitary-adrenal hormonal axis. In addition, only daytime HRV
values were different from baseline measurements, while nighttime measurements only
differ within the isolation period (7).
Several factors may account for the observed results. Operational demands or social
and recreational activities could cause sleep disturbances in spaceflights (5) that in turn
may disrupt other circadian physiological rhythms (14). However, poor sleep quality is
associated with daytime reductions in HF HRV and heartbeat intervals (12) that contrast
with the present observations. Moreover, we failed to demonstrate significant differences in
sleep periods length and phase.
Varying light exposure and reduced sensitivity to Zeitgeber strength during space
missions were also associated with circadian rhythms disruptions. It is known that light is
the dominant environmental input affecting rhythms (14). Sympathetic activity during the
day increases with color light temperature (26) and light intensity (27). In this regard, the
loss of sympathetic predominance reported herein may be associated with a prolonged
exposure to the artificial environmental light of the isolation facility, of lower intensity and
color temperature than natural light.
Changes in mood related to confinement stressors like loneliness, boredom or social
stress should also be considered as factors that may explain the observed results. During a
winter in Antarctica it was reported that tension, anxiety, depression, anger and confusion
decline during the first half of the isolation period and increase close to the end of isolation
(17). Psychophysiological data from this pilot study presented by other authors showed
that, although mood tends to decrease until day 77 of isolation, differences within the 105
days of confinement were not significant (19). Also, apart from mean heart rate, no changes
were seen in sleep HRV indexes, which are known to be sensitive markers to psychological
stressors (9). Thus, the effect of mood changes in the observed results, if any, seems to be
Physical training is another factor that may be associated with changes in HRV.
Confinement of almost any kind is associated with decreased physical load, even with
implementation of special exercise programs. However, opposite to what is seen in the
present study, HF HRV is reduced in sedentary subjects when compared to physically
active ones (8).
The differences between confinement and non-confinement days could be magnified
by a combined effect of increasing tension in anticipation of being locked up (T1) plus
increased exposure to environmental and social demands after the confinement (T5). In this
regard, the reduced environmental stimulation and regularly paced, quite predictable and
well-structured activity schedules during the confinement could be interpreted as a less
stressful and healthier situation (4). However, increased vagal tone by itself should not
15 Download full-text
necessarily be considered as an index of increased health, since it can also be seen in
pathologic conditions such as panic disorder (15).
Although Holter measurements were performed almost exclusively on day-shifts, the
effect of the night shifts in ANS activity should be taken into consideration. When comparing
with day workers, shift work is associated with an increase in sympathetic activity either
during wake periods (increase in heart rate and LF%, decrease in SDNN) or during sleep
periods (decrease in SDNN) (24). Therefore, the increase of parasympathetic activity
associated with long term confinement reported herein is unlikely to be due to the rotating
shift work regime. The 520-day study will provide more data to analyze if the present results
are modified by taking into account the day-time sleep periods associated with night-shifts.
The results from the present study may be important since performance is
associated to autonomic arousal. Specifically, it has been reported that the increase in LF-
HRV and HF-HRV (as seen during confinement) is related to decreased attentional
processing evaluated through an attentional load test (d2 test) (6). Objective physiological
measures could be used to characterize differences in operational efficiency, as well as
abilities to adapt to extreme environments. In turn, fatigue-related performance decrements
caused by sleep loss or sustained operations might be improved with training to regulate
crew-member responses including autonomic and central nervous system parameters (3).
Several limitations should be considered. First, conclusions are restricted due to the
small numbers of subjects, typical of this kind of research. Also, it remains to be seen how
these results would translate to real space missions, where there is always the possibility of