Re: 한번의 격렬한 exercise 서파 안정성 증가시켜 깊은 수면 유도 2021 nature

[문형철]

1. nature  
2. scientific reports  
3. articles  
4. article

Exercise improves the quality of slow-wave sleep by increasing slow-wave stability

Download PDF

• Article
• Open access
• Published: 24 February 2021

Exercise improves the quality of slow-wave sleep by increasing slow-wave stability

• Insung Park, 
• Javier Díaz, 
• Sumire Matsumoto, 
• Kaito Iwayama, 
• Yoshiharu Nabekura, 
• Hitomi Ogata, 
• Momoko Kayaba, 
• Atsushi Aoyagi, 
• Katsuhiko Yajima, 
• Makoto Satoh, 
• Kumpei Tokuyama & 
• Kaspar E. Vogt 

Scientific Reports volume 11, Article number: 4410 (2021) Cite this article

•  
•  
•  
•  

Abstract

Exercise can improve sleep by reducing sleep latency and increasing slow-wave sleep (SWS). Some studies, however, report adverse effects of exercise on sleep architecture, possibly due to a wide variety of experimental conditions used. We examined the effect of exercise on quality of sleep using standardized exercise parameters and novel analytical methods. In a cross-over intervention study we examined the effect of 60 min of vigorous exercise at 60% V˙O2max on the metabolic state, assessed by core body temperature and indirect calorimetry, and on sleep quality during subsequent sleep, assessed by self-reported quality of sleep and polysomnography. In a novel approach, envelope analysis was performed to assess SWS stability. Exercise increased energy expenditure throughout the following sleep phase. The subjective assessment of sleep quality was not improved by exercise. Polysomnography revealed a shorter rapid eye movement latency and reduced time spent in SWS. Detailed analysis of the sleep electro-encephalogram showed significantly increased delta power in SWS (N3) together with increased SWS stability in early sleep phases, based on delta wave envelope analysis. Although vigorous exercise does not lead to a subjective improvement in sleep quality, sleep function is improved on the basis of its effect on objective EEG parameters.

요약

운동은 

수면 지연 sleep latency 시간을 줄이고 

서파 수면(SWS)을 증가시켜 

수면을 개선할 수 있습니다. 

그러나 일부 연구에서는 

운동이 수면 구조에 부정적인 영향을 미친다고 보고하는데, 

이는 다양한 실험 조건이 사용되었기 때문일 수 있습니다. 

저희는 

표준화된 운동 매개변수와 새로운 분석 방법을 사용하여 

운동이 수면의 질에 미치는 영향을 조사했습니다. 

교차 중재 연구에서는 

최대산소섭취량 60%에서 60분간 격렬한 운동을 하면 

심부 체온과 간접 열량 측정으로 평가한 대사 상태와 

자가 보고 수면의 질 및 수면다원검사로 평가한 

후속 수면 중 수면의 질에 미치는 영향을 조사했습니다. 

새로운 접근 방식인 엔벨로프 분석은 SWS 안정성을 평가하기 위해 수행되었습니다. 

운동은 

다음 수면 단계 동안 에너지 소비를 증가시켰습니다. 

수면의 질에 대한 

주관적인 평가는 

운동으로 개선되지 않았습니다. 

수면다원검사를 통해 

빠른 안구 운동 지연 시간이 짧아지고 

SWS에 머무는 시간이 줄어든 것으로 나타났습니다. 

shorter rapid eye movement latency and

reduced time spent in SWS

수면 뇌파를 자세히 분석한 결과, 델타파 엔벨로프 분석에 따르면 

초기 수면 단계에서 SWS(N3)의 델타 파워가 크게 증가했으며 

SWS의 안정성이 증가했습니다. 

격렬한 운동이 

주관적인 수면의 질 개선으로 이어지지는 않지만, 

객관적인 뇌파 파라미터에 미치는 영향을 바탕으로 

수면 기능이 개선됩니다.

Similar content being viewed by others

The Effects of Presleep Slow Breathing and Music Listening on Polysomnographic Sleep Measures – a pilot trial

Article Open access04 May 2020

Role of physical activity and fitness on sleep in sedentary middle-aged adults: the FIT-AGEING study

Article Open access12 January 2021

Bidirectional associations between physical activity and sleep in older adults: a multilevel analysis using polysomnography

Article Open access13 September 2022

Introduction

Epidemiologic studies indicate that insufficient sleep and/or poor sleep quality are associated with multiple adverse effects on health, such as an increased risk for hypertension, type 2 diabetes, and obesity1,2,3,4. Insufficient sleep is also associated with anxiety, depression, and an increased risk for other psychiatric disorders5,6,7. Physical exercise is recommended by academic sleep associations as a low-cost, easily administered, and non-pharmacologic intervention for improving sleep8,9,10,11. A number of studies have demonstrated that a single bout of exercise can decrease sleep onset latency and wake after sleep onset while simultaneously increasing sleep efficiency and slow-wave sleep (SWS)12,13,14,15. Some studies also report that repeated exercise can induce more salient, chronic effects on the sleep architecture14,15. Other studies, however, report few, or even adverse, effects of exercise on the sleep architecture. In healthy young participants, SWS duration was decreased by moderate exercise with an intensity of 35%-45% of maximal oxygen consumption (V˙O2max)16,17. Another study reported no significant differences in the total sleep time and SWS in healthy young men and women exercising at 45%, 55%, 65%, or 75% of the V˙O2max compared to a trial without exercise18. Yet another study reported that 12 weeks of exercise training did not alter the duration of SWS and sleep latency in young female participants19. Although several investigators have attempted to explain these discrepancies by examining differences in experimental protocols such as the sex, age, and exercise habits of the participants, and in the exercise regimen (type, intensity, duration of exercise, and time of day to exercise), the discrepancies in the effects of exercise on sleep remain to be fully explained.

소개

역학 연구에 따르면

수면 부족 및/또는 수면의 질 저하는

고혈압, 제2형 당뇨병, 비만 위험 증가와 같은

건강에 미치는 여러 가지 부정적인 영향과 관련이 있는 것으로 나타났습니다1,2,3,4.

불충분한 수면은

불안, 우울증 및 기타 정신과적 장애의 위험 증가와도 관련이 있습니다5,6,7.

수면 관련 학회에서는

수면 개선을 위한 저비용의 간편한 비약물적 개입으로

신체 운동을 권장하고 있습니다8,9,10,11.

여러 연구에 따르면

한 번의 운동으로

수면 시작 지연 시간과 수면 시작 후 각성 시간을 줄이는 동시에

수면 효율과 서파 수면(SWS)12,13,14,15을 증가시킬 수 있는 것으로 나타났습니다.

일부 연구에서는

반복적인 운동이

수면 구조에 더 두드러지고

만성적인 효과를 유발할 수 있다고 보고하기도 합니다14,15.

그러나

다른 연구에서는

운동이 수면 구조에 미치는 영향이 거의 없거나

심지어 부정적이라고 보고하기도 합니다.

건강한 젊은 참가자의 경우,

최대 산소 소비량(V˙O2max)의 35%~45% 강도의 적당한 운동으로

SWS 지속 시간이 감소한 것으로나타났습니다16,17.

또 다른 연구에서는

건강한 젊은 남성과 여성이

V˙O2max의 45%, 55%, 65%, 75% 강도로 운동했을 때

운동을 하지 않은 시험과 비교했을 때

총 수면 시간과 SWS에 큰 차이가 없었다고 보고했습니다18.

또 다른 연구에서는 12주간의 운동 훈련이 젊은 여성 참가자의 SWS 및 수면 대기 시간에 영향을 미치지 않았다고 보고했습니다19. 여러 연구자들이 참가자의 성별, 나이, 운동 습관, 운동 요법(운동의 종류, 강도, 운동 시간 및 운동 시간)과 같은 실험 프로토콜의 차이를 조사하여 이러한 차이를 설명하려고 시도했지만, 운동이 수면에 미치는 영향의 차이는 아직 완전히 설명할 수 있는 수준은 아닙니다.

For more than half a century, since 1968, sleep has been eval‎uated by applying standardized scoring criteria to electroencephalogram (EEG) and electromyogram recordings established by Rechtschaffen and Kales20. We hypothesized that the discrepancies in the effect of exercise on sleep architecture may at least in part originate from the semi-quantitative nature of this sleep-stage scoring. For example, an epoch is scored as sleep stage N3 or SWS when slow-wave EEG (0.5–2 Hz) with an amplitude greater than 75 µV is observed for more than 20% of a 30-s epoch. A further increase in the amplitude or duration of EEG slow waves does not affect sleep scoring, thereby potentially masking meaningful effects.

As a more quantitative approach to determining sleep depth and quality21, the delta (δ) power (typically 0.5–4 Hz) of the EEG is eval‎uated using fast-Fourier transformation22. Studies assessing the effects of exercise on EEG δ power have produced mixed results. A number of studies demonstrated that exercise is associated with an increase in δ power during subsequent sleep23,24. Young, fit participants also exhibited increased δ power (0–3.9 Hz) after a 30- or 42-km cross-country running race23. In addition, a recent study showed that δ power (0.5–4 Hz) was increased by a moderate (40% of V˙O2max) bicycle ergometer workout in healthy male participants24. In another study, trained athletes who exercised daily at moderate to high intensity were requested to remain sedentary in the laboratory for an entire day, and investigators found no significant differences in the δ power (0.33–3 Hz) between the exercise and sedentary days25.

1968년 이후 반세기 이상,

수면은 Rechtschaffen과 Kales20에 의해 확립된 뇌파(EEG) 및 근전도 기록에

표준화된 점수 기준을 적용하여 평가되어 왔습니다.

우리는

운동이 수면 구조에 미치는 영향의 불일치가

적어도 부분적으로는 이 수면 단계 점수의 반정량적 특성에서

비롯된 것일 수 있다는 가설을 세웠습니다.

예를 들어,

진폭이 75µV보다 큰 서파 뇌파(0.5-2Hz)가 30초 동안 20% 이상 관찰되면

수면 단계 N3 또는 SWS로 점수가 매겨집니다.

뇌파 서파의 진폭이나 지속 시간이 더 증가해도

수면 점수에 영향을 미치지 않으므로 의미 있는 효과를 가릴 수 있습니다.

수면의 깊이와 질을 결정하기 위한

보다 정량적인 접근 방식21으로,

빠른 푸리에 변환22을 사용하여

뇌파의 델타(δ) 파워(일반적으로 0.5-4Hz)를 평가합니다.

운동이 뇌파 δ 파워에 미치는 영향을 평가한 연구 결과는 상반된 결과를 낳았습니다.

여러 연구에 따르면

운동은

이후 수면 중 δ 파워의 증가와 관련이 있는 것으로 나타났습니다23,24.

젊고 건강한 참가자들은

30km 또는 42km 크로스컨트리 달리기 경주 후

δ 파워(0~3.9Hz)가 증가한 것으로 나타났습니다23.

또한,

최근 연구에 따르면 건강한 남성 참가자가

자전거 에르고미터 운동을 적당히(V˙O2max의 40%) 하면

δ 파워(0.5-4 Hz)가 증가한다고 합니다24.

또 다른 연구에서는

매일 중등도에서 고강도로 운동하는 훈련된 운동선수에게

하루 종일 실험실에서 앉아 있도록 요청했는데,

연구자들은 운동하는 날과 앉아있는 날 사이에 δ 파워(0.33-3Hz)에 큰 차이가 없다는 것을 발견했습니다25.

A novel computational method for analyzing EEG waves based on envelope analysis was proposed in 201826. The envelope of a signal in a given frequency band, obtained through its Hilbert transformation, can be viewed as a representation of the instantaneous power in this band. The coefficient of variation of this measure shows how much this power varies over time. The coefficient of variation of the envelope (CVE) thus provides a scale-independent measure of the temporal stability of an oscillation. Low CVE values are found for stable sinusoidal oscillations, intermediate CVE values indicate Gaussian oscillations, and high CVE values are a sign of irregular phasic processes26. We used CVE analysis as a novel tool to investigate the effect of exercise on sleep to examine not only the power of the EEG δ waves generated, but also the stability of these waves.

The present study eval‎uated the effects of a single bout of vigorous exercise in young healthy men on the metabolic state of subsequent sleep and its quality. We wanted to determine, whether exercise improved or decreased sleep quality and whether short exercise bouts can exert lasting effects on the metabolic state.

2018년에 포락선 분석을 기반으로

뇌파를 분석하는 새로운 계산 방법이 제안되었습니다26.

힐버트 변환을 통해 얻은 특정 주파수 대역의 신호 포락선은 이 대역의 순간 전력을 표현한 것으로 볼 수 있습니다. 이 측정값의 변동 계수는 이 전력이 시간에 따라 얼마나 변하는지를 보여줍니다. 따라서 엔벨로프 변동 계수(CVE)는 진동의 시간적 안정성에 대한 스케일 독립적인 측정값을 제공합니다. 낮은 CVE 값은 안정적인 정현파 진동의 경우, 중간 CVE 값은 가우시안 진동을 나타내며, 높은 CVE 값은 불규칙한 위상 프로세스의 징후입니다26. 우리는 운동이 수면에 미치는 영향을 조사하기 위한 새로운 도구로 CVE 분석을 사용하여 생성된 뇌파 δ 파의 힘뿐만 아니라 이러한 파의 안정성도 조사했습니다.

본 연구에서는

건강한 젊은 남성을 대상으로 한 번의 격렬한 운동이 이후

수면의 신진대사 상태와 수면의 질에 미치는 영향을 평가했습니다.

운동이

수면의 질을 향상시키는지 또는 저하시키는지,

그리고 짧은 운동이 대사 상태에 지속적인 영향을 미칠 수 있는지 확인하고자 했습니다.

Results

Participant characteristics

The participant characteristics were (mean ± SEM): age 23.8 ± 0.7 years, weight 66.6 ± 2.2 kg, body fat 17.6 ± 0.01%, and BMI 22.8 ± 0.6 kg/m2. The average V˙O2max was 55.27 ± 5.29 ml/kg/min. All participants completed 2 trials, and there were no significant differences in weight, body fat, and BMI among the trials. All participants fulfilled all inclusion/exclusion criteria (Fig. 1).

Figure 1

Study protocol. The schedule of the control day (upper bar) and exercise day (bottom bar). For participants whose habitual bedtime is at 00:00, indirect calorimetry begins at 11:00 and ends at 08:00 of the next morning, as shown by the dotted rectangles. Participants exited the metabolic chamber at 19:00 for preparation of the polysomnographic measurement and reentered at 21:00. Gray, red, and white boxes represent sleep (00:00–08:00), exercise (17:00–18:00), and wakefulness (08:00–24:00), respectively. Breakfast, lunch, and dinner are denoted by B, L, and D, respectively.

Full size image

Lasting effects on metabolic state

As expected, energy expenditure increased during the exercise period (control trial: 88 ± 3 kcal/h vs. exercise trial: 676 ± 25 kcal/h, p < 0.001; Fig. 2A). As a consequence, oxygen consumption during exercise increased up to 747% (control trial: 0.30 ± 0.01 L/min vs. exercise trial: 2.27 ± 0.08 L/min), HR increased by 238% (control trial: 65 ± 3 beats/min vs. exercise trial: 154 ± 4 beats/min), and core body temperature increased by 0.70 °C (control trial: 36.91 ± 0.07 °C vs. exercise trial: 37.61 ± 0.11 °C) above the sedentary condition. Hourly means of the core body temperature during exercise and 1 h post-exercise were also higher in the exercise trials compared with the control trial. A 2-factor repeated measures ANOVA identified a significant effect of time (p < 0.0001) and interaction (p < 0.0001), although the main effects of group were not significant (Fig. 2B).

Figure 2

Time-course of energy expenditure and core body temperature. Time-course of energy expenditure (A) and core body temperature (B) during the entire experiment is shown. Hourly means ± SE are shown for control (filled black circle) and exercise trials (filled red circle), respectively. The red bar at the bottom represents exercise or a sedentary period, and the gray area represents the sleep period. To attach PSG electrodes, participants exited from the metabolic chamber (19:00–21:00). *Represents a statistically significant difference between control and exercise trials by post hoc comparisons using Bonferroni’s correction for multiple comparisons (*p < 0.05).

Full size image

The mean core body temperature throughout the post-exercise sleep period was not significantly different from that during the control trials (control trials: 36.35 ± 0.06 °C vs. exercise trials: 36.28 ± 0.04 °C, p = 0.40). The hourly core body temperature curves during the sleep period differed between the 2 conditions. A 2-factor repeated measures ANOVA showed no effect of exercise (p = 0.4119), but a significant effect of time (p < 0.0001) and a significant interaction between exercise and time (p = 0.007; Fig. 2B). Energy expenditure remained elevated throughout sleep after exercise (control trial: 526 ± 15 kcal/8 h vs. exercise trial: 544 ± 17 kcal/8 h, p < 0.05; Fig. 2A). Thus, even several hours after a bout of vigorous exercise, the metabolic state was altered in subsequent sleep.

Subjective assessment of sleep quality

Subjective sleep quality on the basis of responses to the OSA-MA questionnaire differed for 'Refreshness' and 'Frequent Dreaming or Nightmares' between the exercise and control conditions, with no significant differences in the other parameters ('Sleepiness on Rising', 'Initiation and Maintenance of Sleep', and 'Sleep Length'; Table 1). Thus vigorous exercise did not improve the subjective assessment of the sleep quality.

Table 1 Subjective parameters by OSA sleep inventory MA version (mean ± standard error).

Full size table

Objective assessment of sleep quality

Basic sleep architecture (i.e., durations of stage 1, stage 2, SWS, REM, and wakefulness after sleep onset) was largely unchanged between the conditions; with the exception of REM, SWS sleep latency, and SWS duration, which were shorter following exercise (Table 2 and Fig. 3). Shortened SWS durations were limited to the first sleep cycle (64.39 ± 5.65 vs. 48.61 ± 3.73 min for control and exercise trials, p = 0.019). SWS episode durations in subsequent sleep cycles were not significantly different (20.28 ± 2.42 min vs. 26.17 ± 4.00 min for control and exercise trials during the second cycle, p = 0.176; 10.17 ± 2.79 min vs. 10.00 ± 2.72 min for control and exercise trials during the third cycle, p = 0.780). At first glance, these results indicate a decrease in slow wave activity. To further eval‎uate this finding, we investigated the power of the δ oscillations in detail. Overall mean δ power throughout the whole sleep period (control trials: 83.67 ± 10.85 μV2 vs. exercise trials: 86.88 ± 9.54 μV2, p = 0.425) was not significantly different between conditions (Fig. 4A). Interestingly, however, δ power in SWS (N3) was significantly larger in the exercise condition (108.4 ± 13.9 μV2) than in the control condition (92.0 ± 14.6 μV2; p = 0.047). Mean δ power in N1 (45.1 ± 9.3 μV2 vs. 41.2 ± 6.2 μV2 for control and exercise trials, p = 0.645) and N2 (51.2 ± 7.1 μV2 vs. 52.6 ± 6.8 μV2 for control and exercise trials, p = 0.711) was similar between conditions. As a consequence, δ wave energy generated over the shortened SWS period was actually larger in the exercise condition compared with the control condition (Fig. 4B–D).

Table 2 Sleep Architecture (mean ± standard error).

Full size table

Figure 3

Time-course of sleep architecture and timing of sleep cycles. (A) Sleep architecture of the 9 participants for the control (upper panel) and exercise trials (bottom panel). Percentage of participants in stage W (wakefulness; black), stage N1 (gray), stage N2 (light blue), SWS (dark blue), and stage REM (red) changed with the sleep time. B and C: Latencies of SWS and REM sleep eval‎uated as time after beginning of sleep cycle (B) and as time after sleep onset (C) are shown. Latency of sleep stage transition in each sleep cycle is shown with black and red box-whisker plots for control and exercise trials, respectively. * and † represent statistically significant differences between the control trial and exercise trial by a paired t-test (*p < 0.05; †p < 0.1).

Full size image

Figure 4

Time-Course of δ-Power of the non-REM Sleep EEG & Relative Occurrence of δ-Power in Each non-REM stage. (A) The 30-min means ± SE of δ-power of the 9 participants are shown as a line graph and accumulated δ-power during non-REM is shown as a bar graph. *Represents a statistically significant difference between the control trial and exercise trial by post hoc comparisons using Bonferroni’s correction for multiple comparisons (p < 0.05). (B–D) Relative occurrences of δ-power in N1 (B), N2 (C), and SWS (D) stages are shown. Inserted bar graphs in each panel represent mean δ-power in each non-REM stage. Black plots (filled black circle) and bars (filled black square) represent control trials, and red plots (filled red circle) and bars (filled red square) represent exercise trials. §Represents a statistically significant difference between control trial and exercise trial by a paired t-test (§p < 0.05).

Full size image

We also performed a detailed examination of the time course of δ wave power throughout sleep. A 2-factor repeated measures ANOVA revealed a significant effect of time (p < 0.0001) and a significant interaction between time and exercise condition (p = 0.0198). Post-hoc analysis showed a significant difference in SWS δ wave-power during 00:30–01:00 after sleep in the exercise condition (Fig. 4A). In summary, we found that the generation of δ wave-power was significantly increased in early sleep phases, without an overall increase in EEG δ wave-power throughout sleep in the exercise trials.

We subsequently examined the stability of the EEG δ waves using CVE analysis. Low CVE values indicate stable, rhythmic, δ wave oscillations, whereas high CVE values indicate short phasic events in the δ frequency range. In an animal model it has been shown that δ-band CVE converges towards 1 as its minimal possible value (see discussion for details). Here we show that in humans this limit holds and δ-band CVE diminishes with increasing sleep depth (CVE of N1 > CVE of N2 > CVE of SWS; Table 3). Our detailed analysis of SWS revealed a significant effect of time (p < 0.0001) and a significant interaction of time and exercise condition (p = 0.0265; Fig. 5A). Post-hoc comparisons showed significant differences in multiple comparisons. Specifically, exercise trials were associated with lower CVE values than control trials in the first half of sleep (1.50 ± 0.03 vs. 1.44 ± 0.03 for control and exercise trials, p = 0.0051; Fig. 5B). This finding reinforces the notion of increased density and stability of δ wave oscillations in early sleep phases after exercise.

Table 3 CVE Values in Each non-REM stage (mean ± standard error).

Full size table

Figure 5

Envelope analysis. (A) Time-course of the CVE during the entire sleep. The 30-min means ± SE of the CVE are shown for the control trial (filled black circle) and exercise trial (filled red circle). (B) Mean CVE during the first half and second half of sleep are shown. Mean CVE is shown for the control trial (open black square) and exercise trial (open red square). Dotted lines connect the same participants. *Represents a significant difference between the control trial and exercise trial by a paired t-test (*p < 0.05). Note that the CVE values did not differ significantly between control and exercise in the last hour of sleep. CVE values were most likely affected by the very low δ power values during this time.

Full size image

Discussion

The present study investigated the acute effects of a single bout of high-intensity exercise on the subsequent sleep phase, as assessed by observation of the metabolic state, responses to a sleep questionnaire, sleep-stage scoring, EEG spectral analysis, and envelope (CVE) analysis of the EEG δ wave band. The parameters of the single 1-h bout of vigorous exercise chosen here were comparable with those of exercises used in studies registering positive effects of exercise on sleep12,13,15 and represent a realistic exercise regimen for healthy adults.

One potential limitation of the study should be considered in the interpretation of the findings. Although not mentioned by the participants, stress due to the unfamiliar sleeping conditions may have affected sleep quality. It should be noted, however, that participants underwent an adaptation day before the experiment and the high sleep efficiency observed in both trials excludes disturbed sleep under the experimental conditions. Moreover, first-night effects would be expected to affect both trial conditions equally owing to the crossover design. To generalize the effects of exercise on sleep, future studies utilizing a different experimental design are warranted, including experiments with a larger sample size. Protocols using regular, chronic exercise with participants of different fitness levels should also be performed.

In the present study, 1 h of vigorous exercise in the evening in untrained volunteers had a moderate, but statistically significant effect on the metabolic state throughout the subsequent sleep phase, detected as excess post-exercise oxygen consumption. Other studies, however, showed that a single bout of low- or high-intensity exercise before lunch did not affect energy expenditure during subsequent sleep27,28. Excess post-exercise oxygen consumption can be interpreted as restoring an oxygen deficit incurred during exercise and more complex mechanisms, including factors that directly (e.g., availability of metabolites such as ADP, ATP, inorganic phosphate, and creatine phosphate) or indirectly (e.g., release of catecholamines, thyroxine, glucocorticoids, fatty acids, calcium ions, and temperature [Q10 effect]) affect mitochondrial O2 consumption29. Interestingly, the increase in energy expenditure in sleep after exercise was not accompanied by an increase in the core body temperature, suggesting an important difference in heat dissipation, which was also observed in previous studies24.

Post-exercise sleep was judged subjectively worse compared to sleep following non-exercise conditions. We hypothesize that mechanisms underlying the post-exercise oxygen deficit and excess oxygen consumption indicate that subjects are under stress and this might explain this lower subjective assessment regarding the 'Refreshness' category in the exercise condition. Another potential reason for subjectively worse sleep after vigorous exercise is muscle soreness as the participants were not accustomed to vigorous exercise. Indeed, in a previous study of moderate (as opposed to vigorous) exercise (~ 45% V˙O2max) in young healthy males, participants reported increased subjective sleep quality, particularly ‘Initiation and maintenance of sleep’30. Recommendations for exercise for non-pharmacologic improvement of subjective sleep quality may benefit from suggestions to participate in moderate exercise, at least initially.

While vigorous exercise may be judged as negatively affecting subjective sleep quality by participants, we found that objective measures of sleep quality indicate a more complex picture suggesting an opposite, beneficial effect. Sleep staging according to American Academy of Sleep Medicine criteria revealed little difference between the exercise and control conditions, consistent with previous studies16,17,18,19. Sleep staging is inherently semi-quantitative, e.g., when the criteria for the SWS stage are fulfilled, further increases in sleep depth cannot be resolved. Notably, stage N4, which might allow for more fine-grained classification of sleep depth, was recently abolished. We observed shortening of the first N3 episode, while several studies investigating the effect of exercise on sleep observed little effect on the total duration of SWS16,17,18,19,31. The participants in our study were not regularly exercising at the level used in this study, which might explain some differences between our findings and those of previous studies, but a recent meta-analysis indicated that fitness level does not modulate the effect of exercise on SWS13.

REM latency is chronically shortened in some pathologic conditions, including depression32 and attention-deficit/hyperactivity disorder33, but these are unlikely causes in the present study, which included young healthy participants. In fact, physical exercise is a known beneficial intervention for depression34. The shortened first REM sleep latency we observed in this study can be interpreted as a forward shift of sleep processes following exercise. Latencies of SWS and REM sleep eval‎uated as time after the beginning of the sleep cycle did not differ significantly during the second and third sleep stages (Fig. 3). The latencies were shortened, however, when eval‎uated as time after sleep onset, i.e., indicating a shift forward. A potential caveat of this type of analysis is the necessarily semi-quantitative nature of the scoring system, which may obscure more subtle differences. As a more quantitative measure, energy in the EEG δ power is viewed as the most reliable indicator of sleep-need buildup and resolution35. Accordingly, the increased δ wave energy production in the first SWS period observed here indicates a more rapid reduction of sleep need in the early sleep phases after exercise, reinforcing the notion of more efficient early sleep processes. Thus, exercise could help achieve efficient sleep earlier by more effectively reducing the sleep need during the first SWS episode. A recent study showed that exercise performed in the evening delays the nocturnal melatonin rise, indicating an effect on the central clock36. Our finding of an advance in the sleep cycle after exercise shows that this mechanism is not responsible for our results.

The lack of an increase in overall δ wave-power throughout the entire sleep period shows that the overall sleep need was not increased by 1 h of vigorous physical exercise. This finding is in contrast to the previous report that sleep after high-intensity exercise (50–70% V˙O2max) increased sleep need as defined by enhanced SWS duration31, but is consistent with findings from a study reporting no effect of high-intensity exercise (65% and 75% V˙O2max) on sleep need18.

Complementary to measuring spectral power, envelope analysis, operating in the time domain, allows an even more detailed analysis of EEG activity, providing information about the morphology of slow waves. According to a recent model based on rat EEG recordings, δ waves originate from the superposition of transient events whose density controls the phasic or continuous appearance of the resulting wave as well as its amplitude26. A few transients (i.e. isolated slow waves) over the EEG background are reported by CVE ≫ 1, while epochs showing dense slow waves are characterized by CVE approaching 1 (for simplicity, CVE = 1 represents a theoretical constant related to Gaussian waves, i.e. the limit for a random superposition of high-density transient events)26. In this study we show that in humans NREM stages are arranged on the CVE scale as 1 < N3 < N2 < N1 (Table 3), hence CVE can be directly interpreted as sleep depth. These morphologic variations in δ waves can be followed quantitatively using envelope analysis, which provides the investigator with a novel tool for assessing the effect of manipulations on sleep that may otherwise elude detection. As a general observation, deep SWS is accompanied by lower CVE values compared with shallow non-REM sleep. The lower CVE values in SWS that we observed here together with the higher δ wave energy in the first SWS period reinforce the notion that the processes generating slow waves are more efficient after exercise compared with control conditions. To our knowledge this is the first report of exercise exerting such an effect. Further investigation into the mechanisms and consequences of this increased δ wave stability are necessary.

Patients who need or wish to perform vigorous exercise during the day may judge their subsequent sleep as inferior compared to rest. Our results indicate that objective parameters contradict this subjective assessment and may serve to reassure individuals, such as athletes who need or wish to perform at high V˙O2max loads, that, if anything, sleep is improved by their physical exercise.

토론
본 연구에서는 대사 상태 관찰, 수면 설문지에 대한 응답, 수면 단계 점수, 뇌파 스펙트럼 분석, 뇌파 δ파 대역의 엔벨로프(CVE) 분석으로 평가한 고강도 운동 한 번의 시합이 후속 수면 단계에 미치는 급성 영향을 조사했습니다.

여기서 선택한

1시간 동안의 격렬한 운동의 매개변수는

운동이 수면에 긍정적인 영향을 미친다는 연구12,13,15에서 사용된 운동과 유사하며

건강한 성인을 위한 현실적인 운동 요법을 나타냅니다.

연구 결과를 해석할 때 연구의 한 가지 잠재적 한계를 고려해야 합니다. 참가자들이 언급하지는 않았지만, 낯선 수면 환경으로 인한 스트레스가 수면의 질에 영향을 미쳤을 수 있습니다. 그러나 참가자들은 실험 전날 적응 기간을 거쳤으며 두 실험 모두에서 관찰된 높은 수면 효율은 실험 조건에서 수면 장애를 배제한다는 점에 유의해야 합니다. 또한 교차 설계로 인해 첫날밤 효과는 두 실험 조건에 동일하게 영향을 미칠 것으로 예상됩니다. 운동이 수면에 미치는 영향을 일반화하기 위해서는 향후 더 큰 표본 규모의 실험을 포함하여 다른 실험 설계를 활용한 연구가 필요합니다. 또한 다양한 체력 수준을 가진 참가자를 대상으로 규칙적이고 만성적인 운동을 사용하는 프로토콜도 수행해야 합니다.

본 연구에서는 훈련을 받지 않은 지원자가

저녁에 1시간 동안 격렬한 운동을 하면

운동 후 과도한 산소 소비로 감지된 이후

수면 단계의 신진대사 상태에 중간 정도이지만

통계적으로 유의미한 영향을 미쳤습니다.

그러나 다른 연구에서는 점심 식사 전 저강도 또는 고강도 운동을 한 번만 해도 이후 수면 중 에너지 소비에 영향을 미치지 않는 것으로 나타났습니다27,28.

운동 후 과도한 산소 소비는

운동 중 발생한 산소 결핍을 회복하는 것으로 해석할 수 있으며,

미토콘드리아 산소 소비에 직접(예: ADP, ATP, 무기 인산염, 크레아틴 인산염과 같은 대사 산물의 가용성) 또는

간접적으로(예: 카테콜아민, 티록신, 글루코코르티코이드, 지방산, 칼슘 이온, 온도[Q10 효과] 방출) 영향을 미치는 요인 등

보다 복잡한 메커니즘이 작용하는 것으로 해석할 수 있습니다29.

흥미롭게도

운동 후 수면 중 에너지 소비 증가는 심부 체온의 상승을 동반하지 않았으며,

이는 이전 연구에서도 관찰된 열 발산에 중요한 차이가 있음을 시사합니다24.

운동 후 수면은

운동하지 않은 상태에서의 수면에 비해

주관적으로 더 나쁜 것으로 평가되었습니다.

우리는 운동 후 산소 결핍과 과도한 산소 소비의 기저에 있는 메커니즘이 피험자가 스트레스를 받고 있음을 나타내며, 이것이 운동 조건에서 '상쾌함' 범주에 대한 낮은 주관적 평가를 설명할 수 있다고 가설을 세웠습니다.

격렬한 운동 후 주관적으로 수면 상태가 나빠지는 또 다른 잠재적 이유는

참가자들이 격렬한 운동에 익숙하지 않았기 때문에

근육통이 발생하기 때문일 수 있습니다.

실제로 건강한 젊은 남성의 중등도(격렬한 운동이 아닌) 운동(최대 산소 포화도 45% 이하)에 대한 이전 연구에서 참가자들은 주관적인 수면의 질, 특히 '수면의 시작 및 유지'30가 증가했다고 보고했습니다. 비약물적으로 주관적 수면의 질을 개선하기 위해 운동을 권장하는 경우, 적어도 초기에는 적당한 운동에 참여하도록 제안하는 것이 도움이 될 수 있습니다.

격렬한 운동은 참가자의 주관적인 수면의 질에 부정적인 영향을 미치는 것으로 판단될 수 있지만, 수면의 질에 대한 객관적인 측정은 그 반대의 유익한 효과를 시사하는 보다 복잡한 그림을 나타냅니다. 미국수면학회 기준에 따른 수면 단계별 분류는 이전 연구16,17,18,19와 마찬가지로 운동 조건과 대조 조건 간에 거의 차이가 없는 것으로 나타났습니다.

수면 단계는 본질적으로 반정량적이며, 예를 들어 SWS 단계의 기준을 충족하면 수면 깊이가 더 이상 증가하지 않습니다. 특히 수면 깊이를 더 세밀하게 분류할 수 있는 N4 단계는 최근 폐지되었습니다. 운동이 수면에 미치는 영향을 조사한 여러 연구에서 운동이 SWS16,17,18,19,31의 총 지속 시간에 거의 영향을 미치지 않는 반면, 첫 번째 N3 단계의 단축이 관찰되었습니다. 본 연구의 참가자들은 이 연구에 사용된 수준으로 규칙적으로 운동하지 않았기 때문에 본 연구 결과와 이전 연구 결과 사이에 약간의 차이가 있을 수 있지만, 최근 메타 분석에 따르면 체력 수준이 운동이 SWS에 미치는 영향을 조절하지 않는 것으로 나타났습니다13.

우울증32 및 주의력 결핍/과잉 행동 장애33를 포함한 일부 병적 상태에서는

렘수면 잠복기가 만성적으로 단축될 수 있지만,

건강한 젊은 참가자를 대상으로 한 본 연구에서는

이러한 원인이 될 가능성이 낮습니다.

실제로 신체 운동은 우울증에 도움이 되는 것으로 알려져 있습니다34. 이 연구에서 관찰된 첫 번째 렘수면 지연 시간이 짧아진 것은 운동 후 수면 과정이 앞으로 이동하는 것으로 해석할 수 있습니다. 수면 주기 시작 후 시간으로 평가한 SWS와 렘수면의 지연 시간은 두 번째와 세 번째 수면 단계에서 크게 다르지 않았습니다(그림 3). 그러나 수면 시작 후 시간으로 평가했을 때는 지연 시간이 짧아져 수면 주기가 앞으로 이동했음을 나타냅니다. 이러한 유형의 분석에서 잠재적으로 주의해야 할 점은 점수 시스템의 반정량적 특성으로 인해 미묘한 차이를 모호하게 만들 수 있다는 점입니다. 보다 정량적인 척도로서, 뇌파 δ 파워의 에너지는 수면 욕구 축적 및 해결에 대한 가장 신뢰할 수 있는 지표로 간주됩니다35. 따라서 여기서 관찰된 첫 번째 SWS 기간의 δ파 에너지 생성 증가는 운동 후 초기 수면 단계에서 수면 필요성이 더 빠르게 감소하여 초기 수면 과정이 더 효율적이라는 개념을 강화합니다. 따라서 운동은 첫 번째 SWS 기간 동안 수면 필요성을 보다 효과적으로 감소시켜 더 일찍 효율적인 수면을 취하는 데 도움이 될 수 있습니다. 최근 연구에 따르면 저녁에 운동을 하면 야행성 멜라토닌 상승이 지연되어 중앙 시계에 영향을 미치는 것으로 나타났습니다36. 운동 후 수면 주기가 앞당겨지는 것을 발견한 것은 이러한 메커니즘이 연구 결과에 영향을 미치지 않는다는 것을 보여줍니다.

전체 수면 기간 동안 전반적인 δ 파 파워가 증가하지 않았다는 것은 1시간의 격렬한 운동으로 인해 전반적인 수면 필요량이 증가하지 않았음을 보여줍니다. 이 결과는 고강도 운동(50-70% V˙O2max) 후 수면이 SWS 지속 시간 증가에 따라 수면 필요성을 증가시킨다는 이전 보고와는 대조적이지만31, 고강도 운동(65% 및 75% V˙O2max)이 수면 필요성에 미치는 영향이 없다는 연구 결과와도 일치합니다18.

스펙트럼 전력 측정과 보완적으로, 시간 영역에서 작동하는 엔벨로프 분석은 서파의 형태에 대한 정보를 제공하여 뇌파 활동을 더욱 상세하게 분석할 수 있습니다. 쥐의 뇌파 기록을 기반으로 한 최근 모델에 따르면, δ파는 일시적인 이벤트의 중첩에서 발생하며, 그 밀도가 결과 파의 위상 또는 연속적인 모습과 진폭을 제어합니다26. 뇌파 배경에서 몇 개의 과도파(즉, 고립된 저속파)는 CVE ≫ 1로 보고되는 반면, 밀도가 높은 저속파를 보이는 시기는 CVE가 1에 근접하는 것이 특징입니다(간단히 설명하기 위해 CVE = 1은 가우스 파에 관련된 이론적 상수, 즉 고밀도 과도 이벤트의 임의 중첩에 대한 한계를 나타냄)26. 이 연구에서는 사람의 경우 NREM 단계가 CVE 척도에서 1 < N3 < N2 < N1로 배열되어 있으므로(표 3), CVE를 수면 깊이로 직접 해석할 수 있습니다. 이러한 δ파의 형태학적 변화는 엔벨로프 분석을 사용하여 정량적으로 추적할 수 있으며, 이는 조사자에게 감지를 피할 수 있는 수면 조작의 효과를 평가할 수 있는 새로운 도구를 제공합니다. 일반적으로 깊은 수면은 얕은 비렘수면과 비교하여 낮은 CVE 값을 동반합니다. 여기서 관찰된 SWS의 낮은 CVE 값과 첫 번째 SWS 기간의 높은 δ 파동 에너지는 운동 후 느린 파동을 생성하는 과정이 대조 조건에 비해 더 효율적이라는 개념을 뒷받침합니다. 우리가 알기로는 운동이 이러한 효과를 발휘한다는 보고는 이번이 처음입니다. 이러한 δ 파 안정성 증가의 메커니즘과 결과에 대한 추가 조사가 필요합니다.

낮 동안 격렬한 운동이 필요하거나 원하는 환자는 이후 수면을 휴식과 비교하여 열등하다고 판단할 수 있습니다. 우리의 결과는 객관적인 매개 변수가 이러한 주관적인 평가와 모순되며, 높은 V˙O2 최대 부하에서 수행해야하거나 수행하기를 원하는 운동 선수와 같은 개인에게 운동으로 인해 수면이 개선된다는 확신을주는 역할을 할 수 있음을 나타냅니다.

Methods

Participants

Nine healthy young men participated in the study. All participants satisfied the inclusion criteria, as follows: 20–30 years of age, body mass index of 18.0–29.9 (kg/m2), a regular sleep/wake pattern, and regular exercise no more than twice a week. Exclusion criteria for the study participants was determined following previous studies22,41; self-reported sleep problems (Pittsburgh Sleep Quality Index score > 5); shiftwork or transmeridian travel within 1 month before the study; smoking; excessive alcohol intake (> 30 g alcohol/day); ongoing medication for cardiovascular disease, diabetes, hypercholesterolemia, hyperglycemia, or hyperlipidemia; and the use of medications affecting sleep or metabolism. Based on sample size calculation, our 9 participants allow us to observe a significant difference with a paired t-test with 75% power and 5% alpha level. Power analysis was conducted by using G-Power 3.1.9 software. This study was conducted according to the guidelines of the Declaration of Helsinki and all procedures involving human participants were approved by the Ethics Committee of the University of Tsukuba. The study protocol was approved by the University of Tsukuba (approval number: tai-28-52) and registered with Clinical Trials UMIN (ID numbers: UMIN000040428, 31/05/2020). All participants provided written informed consent before study commencement.

Procedures

The present study was a randomized-crossover intervention study. The 2 trials were separated by a washout period of 1 week. All participants performed a graded exercise test comprising submaximal and maximal tests using a treadmill (ORK-7000, Ohtake-Root Kogyo Co., Ltd, Iwate, Japan)37 to determine a workload corresponding to 60% of each individual’s V˙O2max. The test was performed within a month before the first experimental trial. Additionally, the experiment was preceded by an adaptation night in the whole-room metabolic chamber, during which the sensors and electrodes of the polysomnographic recording system were attached to the participants. For 5 days prior to the experiment, participants maintained a constant 8-h sleep/16-h wake schedule following their habitual bed and awake time. The participants refrained from ingesting beverages containing caffeine and alcohol, and from performing high-intensity physical activity. Compliance with the instructions was confirmed by sleep diaries and wrist actigraphy (ActiGraph, Ambulatory Monitoring, NY). One day before the experiment and during the experiment day, the participants consumed specified meals at the designated time for breakfast (1 h after waking), lunch (4 h after waking), and dinner (5 h before bedtime).

On the experiment day, the participants arrived at the laboratory, ate lunch, swallowed a core body temperature sensor, and entered the metabolic chamber. The participants performed physical exercise at 60% of the V˙O2max for 60 min beginning at 6 h before bedtime using a treadmill (T1201, Johnson Health Tech Japan, Tokyo, Japan) or remained seated. After the exercise period, the participants were allowed to leave the chamber for 90 min to wipe away sweat and eat dinner. After fitting the participants with the electrodes for polysomnography, they entered the metabolic chamber and remained sedentary. The participants went to bed at their usual bedtime (23:30 ~ 24:30) and slept for 8 h. Energy metabolism was measured for 16 h (from lunch to the next morning; Fig. 1).

The specified meals provided were based on energy requirements estimated from the basal metabolic rate38 with a physical activity level of 1.3 on the day prior to the experiment and control trials. The physical activity level of the exercise trial was assumed to be 1.64 to maintain a stable energy balance39. The macronutrient composition of the meals was 15% protein, 25% fat, and 60% carbohydrates.

MeasuresIndirect calorimetry

The airtight metabolic chamber measured 2.00 × 3.45 × 2.10 m (FHC-15S, Fuji Medical Science Co., Ltd., Chiba, Japan), and air in the chamber was pumped out at a rate of 80 L/min. The temperature and relative humidity of the incoming fresh air were controlled at 25 °C and 55%, respectively. The chamber was furnished with an adjustable hospital bed, desk, chair, and toilet. Concentrations of oxygen (O2) and carbon dioxide (CO2) in the outgoing air were measured with high precision by online process mass spectrometry (VG Prima δB; Thermo Electron Co., Winsford, UK). The precision of the mass spectrometry, defined as the standard deviation for continuous measurement of the calibrated gas mixture (O2, 15%; CO2, 5%), was 0.0016% for O2 and 0.0011% for CO2. Every minute, O2 consumption (V˙O2) and CO2 production (V˙CO2) rates were calculated using an algorithm for improved transient response40. Energy expenditure was calculated from V˙O2, V˙CO2, and urinary nitrogen excretion (N), as described previously22,39,41.

Core body temperature

Core body temperature was continuously monitored using an ingestible temperature sensor that wirelessly transmitted the core body temperature to a recorder (CorTemp, HQ Inc, FL, USA). The sensor was accurate to ± 0.1 °C, and was calibrated by immersion in water at a known reference temperature before use and swallowed 4 h before experiment42.

Self-reported quality of sleep

The Pittsburgh Sleep Quality Index was used to assess sleep quality and sleep disorders in the month prior to the experimental procedures. We assessed 7 components: subjective sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction. The scores of the 7 components were summed to produce a total score (range = 0—21). This index was used only in the preselection stage43. The Oguri-Shirakawa-Azumi sleep inventory MA version (OSA-MA) was used to assess subjective sleep quality after waking in the morning44. This questionnaire comprises 16 items with 5 factors (‘Sleepiness on rising’, ‘Initiation and maintenance of sleep’, ‘Frequent dreaming’, ‘Refreshness’, and ‘Sleep length’).

Polysomnography

The recording system (Alice 5, Respironics Inc, Japan) comprised 6 electroencephalography locations (C3-A2, C4-A1, O2-A1, O1-A2, F3-A2, and F4-A1), submental electromyography, and a bilateral electrooculogram. Sleep parameters were categorized at 30-s intervals as wakefulness and stages N1, N2, SWS, and rapid eye movement (REM) sleep according to the standard criteria of the American Academy of Sleep Medicine45. In addition, total sleep time, sleep onset latency, REM sleep latency, and sleep efficiency were eval‎uated.

Data analysis: spectral analysis of the electroencephalogram

The C3-A2 EEG recording was analyzed using discrete fast-Fourier transformation techniques as previously described22. Fast-Fourier transformation was conducted on an EEG record length of 5 s to obtain a frequency resolution of 0.2 Hz. Each 5-s EEG segment was first windowed with a Hanning tapering window prior to computing the power spectra. The spectral distribution was categorized into the following frequency bands: delta (δ: 0.75–4.00 Hz), theta (θ: 4.10–8.00 Hz), alpha (α: 8.10–12.00 Hz), sigma (σ: 12.10–14.00 Hz), and beta (β: 14.10–30.00 Hz)22. The power content of the δ band for each 30-s epoch of sleep was determined as the mean of the δ power measured in six consecutive 5-s segments of the EEG (expressed as μV2).

Envelope analysis

The CVE for the δ band was calculated for EEG recordings (C3-A2) at 30-s intervals. To minimize aliasing effects, the epochs had 50% overlap (i.e., epoch length = 60 s). First, every epoch was digitally bandpass-filtered (0.5–4 Hz) with a fourth-order IIR implementation of a Butterworth filter using the 'signal' package for the R language (http://r-forge.r-project.org/projects/signal/). The envelope of the filtered EEG (filt_EEG_envelope) was obtained using its Hilbert transform (Ht) according to the standard relation:

Filt_EEG_envelope = sqrt (filt_EEG2+ Ht(filt_EEG)2),

where sqrt corresponds to the square root. Both the filter and envelope calculations usually produce artifacts at the border of each epoch. To avoid this problem, the samples of each epochs were collected with a 10% excess (i.e., totaling 66 s, 3 s per side). Once the envelope was obtained, this time excess was excised. The mean and standard deviation (SD) of the obtained envelope were calculated and a normalized version of the coefficient of variation (CVE) was obtained sd/(mean*0.523); with 0.523 being the value for Gaussian waves. As a consequence, CVE values larger than 1 result from processes more phasic than Gaussian waves, while values below 1 indicate more sinusoidal processes. For each epoch, the coefficient of variation (i.e. SD/mean) of the corresponding envelope was stored as a relevant feature26.

Statistical analysis

The results are expressed as the mean ± standard error of the mean (SEM). Paired Student's t tests were used to compare the total amount of δ power during the whole sleep period, each sleep stage latency, the OSA-MA parameters, and the sleep parameters between the mean value of trials. The effects of exercise on the time course of δ power, CVE, core body temperature, and energy expenditure were assessed by 2-way repeated-measures analysis of variance (ANOVA) and Bonferroni’s correction for multiple comparisons. 1-way ANOVA and Bonferroni’s correction for multiple comparisons were used to compare the CVE in each non-REM stage. Data analysis was conducted using Prism 8 (GraphPad Software, San Diego, CA), or R (https://www.r-project.org/), and differences were considered significant when the error probability was less than 0.05.

References

1.  
2.  
3.  
4.  
5.  
6.  
7.  
8.  
9.  
10.  
11.  
12.