Heart rate variability rebound following exposure to persistent and repetitive sleep restriction

Abstract

While it is well established that slow-wave sleep electroencephalography (EEG) rebounds following sleep deprivation, very little research has investigated autonomic nervous system recovery. We examined heart rate variability (HRV) and cardiovagal baroreflex sensitivity (BRS) during four blocks of repetitive sleep restriction and sequential nights of recovery sleep. Twenty-one healthy participants completed the 22-day in-hospital protocol. Following three nights of 8-hr sleep, they were assigned to a repetitive sleep restriction condition. Participants had two additional 8-hr recovery sleep periods at the end of the protocol. Sleep EEG, HRV, and BRS were compared for the baseline, the four blocks of sleep restriction, and the second (R2) and third (R3) nocturnal recovery sleep periods following the last sleep restriction block. Within the first hour of each sleep period, vagal activation, as indexed by increase in high frequency (HF; HRV spectrum analysis), showed a rapid increase, reaching its 24-hr peak. HF was more pronounced (rebound) in R2 than during baseline (p < 0.001). The BRS increased within the first hour of sleep and was higher across all sleep restriction blocks and recovery nights (p = 0.039). Rebound rapid eye movement sleep was observed during both R2 and R3 (p = 0.004), whereas slow-wave sleep did not differ between baseline and recovery nights (p > 0.05). Our results indicate that the restoration of autonomic homeostasis requires a time course that includes at least three nights, following an exposure to multiple nights of sleep curtailed to about half the normal nightly amount.

Statement of Significance

In this study, an autonomic rebound (i.e. increased vagal activity) was found to occur during recovery sleep following a novel repetitive sleep restriction challenge. Baroreflex sensitivity activation during sleep increased during repeated exposure to sleep restriction and persisted during the second and third recovery nights of sleep, tempering the possible overshoot of blood pressure dipping. These autonomic changes during recovery sleep reflect the cardiovascular recovery processes that follow the accumulation of sleep deficit, and persist for three nights, following resumption of 8 hr of sleep opportunity per night. The autonomic changes during recovery sleep provide evidence of accumulated autonomic pressure during repetitive sleep restriction, and suggest that sleep is important for homeostatic regulation of cardiovascular function.

Introduction

It is not unusual for workers to restrict sleep during the work days and to try to “catch up” on the days off [1]. Recent epidemiological studies demonstrate a consistent link between sleep deprivation and hypertension [2, 3]. It is not known whether the “catch-up” sleep offsets the adverse effects of restricted sleep on cardiovascular risk markers, or how much sleep is needed to recover from the deficit. This study aimed to investigate the cardiovascular recovery process by measuring indices of cardiovascular function associated with restricting and recovering sleep chronically in daily life.

Mounting evidence suggests that sleep is important for cardiovascular regulation [4]. Normal sleep in healthy humans is accompanied by changes in blood pressure (BP) and heart rate (HR), which are primarily regulated by the autonomic nervous system (ANS), with non-rapid eye movement (NREM) slow-wave sleep (SWS, or N3) characterized by parasympathetic predominance and sympathetic inhibition. BP and HR decrease during NREM sleep, and increase during rapid eye movement (REM) sleep [5]. At this time, however, very little is known about the autonomic modulation in recovery sleep that follows sleep deprivation.

A few studies have shown that sleep that ensues either immediately following a night of acute total sleep loss [6–8], or immediately following several nights of partial sleep deprivation [9, 10], contain rebound amounts of SWS and delta power. REM sleep is initially suppressed, but rebounds in subsequent nocturnal sleep periods [10, 11]. However, to our knowledge, there are no studies that have investigated the prolonged recovery of autonomic indices during second or third recovery sleep that follows sleep deprivation. This study was conducted to address this important gap in knowledge.

Spectral analysis of heart rate variability (HRV) has been widely used as a non-invasive autonomic marker in human studies for years. Impaired balance of ANS assessed by HRV, characterized by parasympathetic withdrawal (decreased normalized high frequency [HF]), has been suggested as underlying pathophysiological mechanisms linking sleep deprivation and development of hypertension and cardiovascular diseases [12–14]. The baroreflex is a rapid negative feedback mechanism that contributes importantly to BP homeostasis. Cardiovagal baroreflex sensitivity (BRS) is an index used to quantify how much control the baroreflex has on the HR, and is estimated by the responses in HR to changes in systolic BP (SBP). Dysfunction of the arterial baroreflex has been linked to hypertension [15, 16]. Furthermore, cardiovagal BRS has been reported to increase during sleep, in comparison to wake periods [17].

In the current investigation, both HRV spectrum analysis and cardiovagal BRS were used to investigate the autonomic indices during repetitive exposure to sleep restriction (SR) and subsequent recovery sleep. In order to investigate longer-term recovery processes, this study investigated sleep-associated recovery of autonomic indices (HRV and BRS) during the second and third night of recovery sleep, following exposure to repetitive blocks of sleep loss. The first night of recovery sleep was not recorded, in order to limit the participant burden (a total of seven 24-hr recordings were performed in the protocol); and also to allow an undisturbed night to precede investigation of further recovery processes. While previous human studies reported strong coupling between delta power and HRV during sleep [18–20], the autonomic recovery time course has not been investigated, to our knowledge, beyond a single night. The current study hypothesized a prolonged autonomic rebound during recovery sleep following repetitive SR.

Methods

Participants

Twenty-one healthy participants who completed the SR condition of a larger study of 45 participants were included in this analysis, consisting of 11 women and 10 men between ages of 21 and 53 years; average age 31 ± 2 years (average body mass index 24 ± 1 kg/m²) [21]. Medical history and physical screening were performed in order to rule out anyone with psychiatric, neurological, pain-related, immune, cardiovascular diseases, or significant allergy before enrolling in the study. All participants wore an actigraph and completed sleep diaries for 2 weeks to ensure that they had regular daily sleep duration between 7 and 9 hr, and habitual sleep periods starting within 1.5 hr of 2300. Furthermore, all participants underwent overnight sleep screenings to rule out sleep disorders (e.g. sleep apnea and restless leg syndrome) before being enrolled for the in-hospital phase of the protocol. The exclusion criteria included respiratory disturbance index of more than 5 events/hr on polysomnographic sleep study, leg movements with arousal more than 10/hr, or sleep efficiency less than 80% (independent criteria). Female participants reported the timing of their menstrual cycle, in order to schedule the start of their study in late luteal phase (four participants) or early follicular phase (six participants) of their menstrual cycle. One female participant was post-menopausal.

Experimental design

This study was approved by the Beth Israel Deaconess Medical Center Institutional Review Board and informed consent was obtained prior to testing for eligibility. The overnight sleep screening and 22-day in-hospital stay phase was carried out in the clinical research center (CRC) at Beth Israel Deaconess Medical Center. Following 2 days of adaptation and 1 day of baseline recordings with 8-hr sleep from 2300 to 0700, participants slept 4 hr/night for 3 consecutive days from 0300 to 0700, followed by one night of 8-hr recovery sleep. Participants repeated this sleep block three more times, then had two additional nights of 8-hr sleeps (see Figure 1). The participants were asked to remain in bed in a semi-supine position with lights dimmed (<20 lux) between 2300 and 0300 during SR nights, in order to minimize possible differences between SR and baseline the same period (2300–0300), other than sleep opportunity. Each participant was accompanied by a research assistant who helped to maintain their wakefulness, with no food or water allowed. The participants were instrumented with Portapres and Polysomnography recording equipment on the baseline, SR (shown in Figure 1; the final deprivation day in each restriction block), R2 (second recovery night following last SR block), and R3 (last recovery night in CRC).

Figure 1.

Study protocol. Following three nights of 8 hr of sleep (from 2300 to 0700), a repetitive four blocks of SR model was introduced with three nights of 4 hr of sleep (from 0300 to 0700), each followed by one night of 8-hr recovery sleep. The participants had two additional 8 hr of sleep at the end of the protocol (R1, first recovery night following last SR block; R2, second recovery night; R3, third recovery night). Twenty-four-hour continuous beat-to-beat BP recordings and polysomnography were performed at baseline, the last restrictive day of each cycle, and the last two recovery nights, as shown in black lines.

Diets designed to maintain participant’s body weight/composition were provided at standard times for breakfast (0730), lunch (1230), and dinner (1830). Fats, carbohydrates, sodium, and potassium were controlled in the designed meals. Water was provided in hourly aliquots based on weight-adjusted norm, and adjusted in the first 2 baseline days to accommodate individual preference. Room temperature was maintained at participant preferred level (adjusted during first 2 baseline days) during the day, and dropped by 2°C at night. Participants’ daily physical activity levels were controlled by hourly walks (with exception of days when participants were instrumented) and use of the hospital fitness center three times per week.

Participants were encouraged to maintain contact with their social network while in the CRC, were permitted visitors, and had access to phone and computer for email, except on heavy recording days. Participants were asked to not use the bed during wake time after getting up in the morning. All participants were supervised and accompanied by a research assistant during wake time to help maintain their wakefulness, safety, and social interaction throughout their whole stay. During the 22-day stay, the participants’ regular sedentary activities were talking, reading, writing, watching TV/movies, and playing video/board games.

Measurements

Actigraph and sleep diary

All participants were given an Actiwatch (Actiwatch 2, Respironics Inc., Murrysville, PA) along with a sleep diary to evaluate their habitual bedtimes and sleep times according to inclusion criteria.

Beat-to-beat BP

A digital photoplethysmography (Portapres, TPI, Brussels, Belgium) was used to record beat-to-beat BP from the arterial waveform in a finger of the non-dominant hand. Two small finger cuffs were applied to fingers of the participant’s non-dominant hand, which continuously inflated and deflated to measure the beat-to-beat BP. The alteration between cuffs was set at 30- or 15-min intervals depending on the participants’ tolerance. A research technician attended throughout the recording to ensure continuous quality of recordings. In addition, warming packs were used as needed to prevent artifacts due to constricted arteries in the digits. This design provides continuous beat-to-beat recordings over longer periods (i.e. up to 24 hr in this study). Portapres measurements have been shown to correlate very well with directly measured beat-to-beat radial artery pressures [22, 23] and have been used extensively to measure 24-hr BP variations [24, 25].

Polysomnographic recording

Sleep was recorded using the Embla system (Natus Medical Incorporated, Pleasanton, CA) on the screening night visit, and at baseline, SR, R2, and R3 during in-hospital stay (Figure 1). The montage followed standard American Academy of Sleep Medicine criteria, and sleep electroencephalography (EEG) was manually stage scored (single scored by one sleep technician) on a 30-s epoch basis [26]. The montage included standard recording sites F3, F4, C3, C4, O1, and O2, referenced to linked mastoids for EEG analysis. Two leads ECG electrodes were placed at sub-clavicle fossa for HR recording and HRV analysis.

Data analysis

Beat-to-beat BP

Beat-to-beat BP measured by the Portapres system was fed directly into the Embla system, ensuring synchronization of signals between BP and HR for use in subsequent BRS analysis. Then the data were transformed to European Data Format and further imported to LabChart (LabChart 7, ADInstruments Co.) for offline analysis. The raw data were first visually inspected by an investigator with the detailed recording notes to exclude non-analyzable/non-reliable data. Customized analysis settings of the LabChart BP module, including minimum peak height of 20 mm Hg (equivalent to pulse pressure), minimum period of 400 ms (equivalent to R-R interval [RRI]), and minimum height of dicrotic notch of 5% of peak height, were used to detect the qualified waveforms. Artifacted waveforms were automatically removed. For each 24-hr recording, BP waveforms with any one of the five parameters (systolic, diastolic, mean arterial pressure, BP pulse pressure, or waveform period), deviating more than 3 SDs (calculated based on all data collected within the 24-hr period), were removed (<5% of data from each recording). In total, approximately 11% beat-to-beat BP data were missed and or removed including missing data due to malfunctioning devices, data excluded from visual inspection, and data deviating more than 3 SDs.

Heart rate variability

After HR signals were collected from ECG recordings by Remlogic (Natus Medical Incorporated), the data were transformed to European data format and imported to LabChart for offline analysis. Waveform measurement was used to detect and analyze the ECG data (HRV module 2.0 LabChart 7, ADInstruments Co.). R wave detection was largely automated; waveforms were identified with customized preset detection settings for each individual recording. An investigator visually inspected the data and fine-tuned in the detection adjustments panel. Artifact signal due to movement and ectopic waveforms were removed from the analysis.

HRV was assessed by power spectral analysis of RRI [27]. Specifically, beat-to-beat RRI tachogram is windowed (Hann window) and linear de-trended. Next, the Lomb–Scargle periodogram algorithm is performed to generate the power spectrum analysis. We analyzed the data in 10-min segments (from the start of each recording), and subsequently averaged into hourly bins. A sub-analysis by Remlogic (Natus Medical Incorporated) was done for analyzing HRV in 5-min segments according to different sleep stages between 2300 and 0000. Specifically, each 5-min segment between 2300 and 0000 was determined as one sleep stage or wake stage (e.g. if the majority of 30-s epochs of sleep stage was stage 2, then this 5 min will be signed as stage 2). Next, the HRV data were averaged for each sleep stage. The magnitude of RRI is quantified by calculating the power spectral density for the signal in low frequency (LF; 0.04–0.15 Hz) and HF (0.15– 0.4 Hz). HF power is predominantly a reflection of vagal activity, whereas both LF power and LF/HF ratio provide gross estimates of sympathovagal balance [27]. Normalized units (i.e. “nu”) of the spectral powers of HF [HFnu = HF/(LF + HF)] will be presented. Due to algebraic redundancy, normalized LF [LFnu = LF/(LF + HF)] and LF:HF ratio show the equivalent information about sympathovagal balance, they will not be presented [28].

BRS from spontaneous method

Spontaneous BRS is determined from beat-to-beat changes in RRI and SBP (sequence method), as originally reported by Bertinieri et al. [29] and modified by Blaber et al. [30]. Baroreflex sequences are identified by three or more beats relating to RRI and progressive, spontaneous changes of SBP (lag 1). Both up-up sequences (progressive increases of SBP followed by a lengthening of the RRI) and down-down sequences (progressive decreases of SBP with a subsequent shortening of the RRI) were analyzed. Minimum criteria for detecting a sequence was set at 1 mm Hg for SBP and 4 ms for RRI (Custom module, LabChart 7, ADInstruments Co.). Linear regression analysis was used to determine the slope of the linear relationship between the RRI and SBP for each sequence. Only sequences with linear r values more than 0.8 were accepted. Up-up or down-down sequences were then averaged to hourly bins for each participant. Ten participants had 1 day (out of 3 heavy recording days) of missing BRS data (~16%), due to malfunctioning devices (either ECG recording or Portapres).

Delta power analysis

As an index of SWS intensity, delta spectral power band (0.5–4.0 Hz) was expressed in normalized units, which were defined as the ratio between delta power value and the total sum of power spectrum values over all frequencies. Spectral analysis of EEG data was performed on the C3-A2 channel with fast Fourier transformation (FFT) to estimate power spectra. In cases where C3-A2 channel was of insufficient quality, the C4-A1 was used instead. After a linear detrending, FFT was applied using a rectangular window, the data were analyzed in 10-min segments (from the start of each recording), and subsequently averaged into hourly bins. Delta power data are presented as percent change of baseline’s value.

Statistical analysis

All data were analyzed statistically using commercial software (IBM SPSS 22.0; IBM Corp, Armonk, NY). Mixed linear model was used to estimate the effects of repeated SR on both sleep stages and autonomic parameters, during the first hour of sleep throughout blocks of SR and recovery. The fixed factor is SR effect (baseline, one to four blocks of SR, R2, or R3). Assigned participant code was treated as random effect. Pairwise comparison within the mixed linear models was used to estimate at which points values differed in blocks of SR and recovery from baseline; Least significant different test was used to adjust the multiple comparisons. A method reported by Bland and Altman [31] was used to calculate correlation coefficients between SWS and HF within participants. Results are expressed as means ± SE, and p <0.05 level of rejection was used.

Results

The averaged global Pittsburgh Sleep Quality Assessment score measured during screenings was 1.57 ± 0.2. Before entering the study, the participants’ habitual sleep duration was 7.8 ± 0.1 hr and sleep efficiency was 83 ± 1%, estimated by the actigraph and sleep log. The results of whole night sleep and sleep stages during baseline, R2, and R3 are presented in Table 1 (whole night sleep stages during SR have been published) [21]. REM sleep showed a robust rebound during both recovery nights R2 and R3 (SR effect p = 0.004). Minutes of NREM did not change during either of the recovery nights in comparison with the baseline. Wake after sleep onset decreased during R2 and R3 (SR effect p = 0.013) in comparison to the baseline. Furthermore, sleep efficiency was significantly increased in R2 compared to baseline (SR effect p = 0.023).

Table 1.

Whole night NREM and REM sleep time

	Baseline	R2	R3	p value
TST, min	411 ± 6	429 ± 6*	422 ± 6	0.022
REM sleep, min	79 ± 5	97 ± 5*	96 ± 5*	0.004
N1, min	40 ± 4	33 ± 4	35 ± 4	0.069
N2, min	207 ± 6	215 ± 6	213 ± 6	0.485
SWS, min	84 ± 5	84 ± 5	78 ± 5	0.275
SOL, min	20 ± 4	16 ± 4	19 ± 4	0.532
WASO, min	49 ± 4	35 ± 4*	38 ± 4*	0.013
Sleep efficiency, %	86 ± 1	90 ± 1*	88 ± 1	0.023

N1 and N2, stage 1 and 2 non-REM sleep; R2, second recovery night; R3, third recovery night; SOL, sleep onset latency; TST, total sleep time; WASO, wake after sleep onset.

*Significantly different (p < 0.05) compared to baseline in each group.

Figure 2 shows the hourly averaged normalized HF (Figure 2, A) and delta power (Figure 2, B) during sleep at baseline, SR (left panels), and R2 and R3 (right panels). Normalized HF significantly increased within 1 hr of sleep onset during nights of 8-hr sleep (2300–0000, right panel) compared to 1 hr before sleep onset (2200–2300) in the baseline (p = 0.003), R2 (p < 0.001), and R3 (p = 0.019). Similarly, normalized HF also significantly increased within 1 hr of sleep onset during nights of 4-hr sleep (0300–0400, left panel) compared to 1 hr before sleep (0200–0300) throughout all four SR blocks (p < 0.01 for all). Figure 3 depicts the normalized HF, BRS up-up, and BRS down-down sequences within the first hour of sleep. Normalized HF did not change throughout the four blocks of SR (SR effect p = 0.181), but showed a rebound effect during R2 (p < 0.001). Within the first hour of sleep, cardiovagal BRS from up-up sequences did not show significant changes across the four blocks of SR or recovery. In contrast, BRS from down-down sequences significantly increased across four blocks of SR and during R2 and R3 compared to baseline (SR effect p = 0.039).

Figure 2.

Hourly averaged normalized HF from HRV spectrum analysis (A), and delta power from EEG spectrum analysis (B) during four blocks of SR (left panels) and recovery (right panels), with the baseline tracing repeated for ease of comparison. For each hourly bin, the corresponding time across the day on the x-axis represents the time immediately after the end of that hour collection. For example, the bin plotted at 23 on x-axis represents the HF data collected from 2200 to 2259. The blue bar represents the sleep period in 8-hr baseline and recovery sleep opportunities (from 2300 to 0700), whereas the red bar represents the sleep period in 4-hr sleep opportunities (from 0300 to 0700). R2, second recovery; R3, third recovery.

Figure 3.

Normalized HF from HRV spectrum analysis, BRS from up-up (BRS UpUp) and down-down (BRS DD) sequences within first hour of sleep (2300–0000 for 8-hr sleep, 0300–0400 for 4-hr sleep). *Significantly different (p < 0.05) from baseline. R2, second recovery; R3, third recovery.

SBP and HR within the first hour of sleep did not change significantly across the protocol (SR effect p = 0.212 and 0.451, respectively). Changes (Δ) of SBP and HR from wake to sleep (calculated using the average values of 1 hr before sleep minus the values within 1 hr of sleep) during 8-hr sleep nights (i.e. baseline, R2, and R3) are plotted in Figure 4 (upper panel). The sleep-induced reduction of SBP did not show significant differences in R2 or R3 compared to baseline (SR effect p = 0.215). However, sleep-induced reduction of HR was significantly exaggerated in R2 compared to baseline (SR effect p = 0.033). Further post hoc analysis of sleep (Figure 4, lower panel) within the first hour found that normalized HF was significantly elevated during R2 for N2 (p = 0.040) and SWS (p = 0.041) compared to baseline, but not during R3 (p = 0.879 and 0.302 during N2 and SWS, respectively).

Figure 4.

Upper panel: Changes of SBP and HR between 1 hr before sleep (2200–2300) and within 1 hr of sleep (2300–0000) during baseline and recovery nights. Lower panel: Normalized HF during N2 (left) and SWS (right) within 1 hr of sleep during baseline and recovery nights.*Significantly different (p < 0.05) from baseline. R2, second recovery; R3, third recovery.

Figure 5 depicts the delta power and SWS within the first hour of sleep during baseline, throughout the four blocks of SR, R2, and R3. Delta power at baseline was set as 100% in Figure 5, the percent change of delta power from baseline significantly increased throughout the four blocks of SR and R2 (SR effect p = 0.007). SWS was significantly increased during the first three blocks of SR (SR effect p = 0.006).

Figure 5.

Delta power from EEG spectrum analysis and SWS within the first hour of sleep (2300–0000 for 8-hr sleep, 0300–0400 for 4-hr sleep). Delta power data are presented as percent change from baseline. *Significantly different (p < 0.05) from baseline. R2, second recovery; R3, third recovery.

Table 2 presents the Bland and Altman correlation coefficients between whole night sleep (SWS or REM) and HF within participants. There were significant positive correlations between SWS and HF (all p < 0.001), and negative correlations between REM and HF (all p < 0.001) in each of the 3 recording days.

Table 2.

Altman correlations of repeated measurements from whole night recordings

	SWS and HF		REM and HF
Study day	R	p	R	p
Baseline	0.44	<0.001	−0.31	<0.001
R2	0.49	<0.001	−0.49	<0.001
R3	0.56	<0.001	−0.49	<0.001

R2 second recovery night; R3, third recovery night.

Discussion

Homeostatic characteristics of sleep EEG in the first night following acute sleep deprivation have been well described, and include increased SWS and delta power, and our findings during short sleep blocks are consistent with this. However, very little research has focused on the recovery of autonomic indices. This is the first study to demonstrate the prolonged recovery process of autonomic indices following persistent and repetitive SR. During the second night of recovery sleep, following repetitive runs (blocks) of persistent exposure to insufficient sleep, HRV showed a rebound in the first hour of sleep, and its HF component was correlated with delta power. This finding suggests a role for autonomic homeostasis associated with the cortical EEG recovery process. Moreover, BRS activation (down-down sequences) tempered the possible overshoot of BP dipping in the early part of the night during R2 and R3, when vagal activation and SWS predominated. This pattern was followed by continued REM rebound later during these recovery sleep nights. These findings support the conclusion that the ongoing autonomic normalization process continues after more than two nights of recovery sleep.

Normal healthy sleep is accompanied by changes in ANS. Normalized HF of HRV is typically increased from awake to NREM sleep and is decreased during REM sleep [19, 32]. Highly controlled experimental studies have shown that sleep deprivation significantly alters the ANS balance. Specifically, acute total sleep deprivation contributed to lower parasympathetic cardiac regulation, assessed by spectral HRV analysis [33]. Partial SR studies of one night [14] and over several consecutive nights [12] have reported consistently impaired sympathovagal balance during wake time following nocturnal SR. Furthermore, following three nights of SWS suppression, daytime normalized HF decreased by 15% and sympathovagal balance (LF/HF ratio) was 14% higher compared to baseline sleep [13]. Studies have reported acute decrease of parasympathetic cardiac regulation during the wakeful daytime period following SR. However, to our knowledge, the autonomic regulation during the recovery sleep following sleep deprivation has not been described.

Previous human studies have reported strong coupling between EEG and HRV during sleep [18–20]. Brandenberger et al. reported that the ultradian oscillation in HRV linked in a “mirror-image” to the overnight oscillation of delta power [18]. Jurysta et al. extended the analysis to all conventional EEG frequency bands, and found that EEG spectral bands other than delta are also closely linked to changes in cardiac autonomic control [19], although the gain of coherencies between HF and EEG power spectra was highest for delta band. The evidence suggests that the brain-stem-directed cardiac autonomic control and sleep are not related to each other by a simple oscillator mechanism [18–20] and there is a time delay between the two physiological systems. The cardiac autonomic control is preceded by several minutes of delta EEG variability during sleep [19]. However, the underlying mechanism of this relationship is not clear. Moreover, in the current study, ECG and EEG were not recorded during the first recovery night because the primary focus was on the longer time course (over subsequent nights) of recovery following repetitive SR exposure.

The significant correlation between hourly bins of SWS and HF (Table 2) reported here provide further support for the coupling of EEG and HRV during sleep. This relationship appeared to be even more pronounced during recovery sleep, supporting a role for sleep in the autonomic recovery process. Moreover, the changes in HRV were apparently more influenced by sleep than by circadian time. During the SR period (2300–0300), the HF was not significantly increased compared to baseline at the same circadian time (Figure 2, A). When sleep was initiated, HF was significantly increased compared to pre-sleep period (i.e. 0200–0300), although it would be expected that circadian influence would have predicted a downregulation of HF at this time.

There is little research that has investigated autonomic parameters during recovery sleep. Glos and colleagues found that after 40 hr of continuous wakefulness, LF/HF ratio was reduced (indicating enhanced HF) along with increased SWS during recovery sleep compared to baseline [34]. The authors suggested that the decrease of sympathovagal balance was due to enhanced sleep consolidation and an increase in SWS [34]. Viola et al. [35] also reported increased slow-wave activity during recovery after 40-hr sleep deprivation, but in contrast to Glos et al. [34], this rebound of slow-wave activity was not accompanied by an increase of parasympathetic activity [35]. In the current study, normalized HF was increased (Figure 3) during R2, and percent change of delta power (Figure 5) was also increased. While the increase in delta power did not generate a significant increase in SWS (Figure 5), our results do indicate that autonomic and sleep homeostasis have not completely recovered. Furthermore, despite that neither BP nor HR during R2 were different from baseline, the sleep-induced reduction of HR was significantly exaggerated in R2 compared to baseline (Figure 4). These results indicate that a longer recovery process, expressed as rebound, for autonomic and EEG homeostasis may be needed.

Recovery following partial sleep deprivation in EEG measurements has been investigated. It has been reported that recovery sleep following curtailment of sleep time to 4 hr per night for four nights leads to increased SWS and REM sleep during the first recovery night, accompanied by a reduction of wake time, N1, and N2 sleep [10]. Moreover, the rebound of REM sleep persisted into the second night of recovery sleep, whereas SWS returned to baseline levels. As noted in our previous paper [21], the circadian placement of the short sleep period between 0300 and 0700 resulted in higher amounts of REM sleep during the first 4 hr of sleep in the restricted condition, compared with the first 4 hr of the 8-hr baseline sleep. Consistent with previous findings, the current study found no rebound of SWS during R2 or R3 (Table 1). However, REM rebound persisted into R2 and R3 in the current study, underscoring the severity of the sleep deficit.

Cardiovagal BRS, an index of autonomic regulation on BP via modulation of HR, has been reported to increase during sleep in comparison to the wake period [36]. Consistent with existing literature, we found increased cardiovagal BRS during the transition from wake to sleep. We also found that BRS was significantly increased within the first hour of sleep during SR and R2 and R3 in comparison to baseline, when analyzed with respect to the downregulation, of BP. During the transition from wake to sleep (i.e. within first hour of sleep), when BP decreased, BRS estimations showed a fast adjustment by increasing HR to prevent further decrease of BP during SR and recovery nights.

The BRS results indicate that SR interrupted the balance of BRS regulation by increasing the downregulation without changes in the upregulation. We speculate that SR may either enhance the deactivation effects of baroreceptors or decrease the afferent signals traveling within the glossopharyngeal and vagus nerves to the brain-stem when resting BP decreases during transition from wake to sleep. The overall efferent autonomic signal (i.e. not separating up- or downregulation of resting BP) was parasympathetic predominance (i.e. along with enhanced delta power) during R2. Moreover, these BRS changes were further maintained during R2 and R3, indicating that one night of recovery sleep is insufficient to restore autonomic homeostasis. The underlying mechanism is unknown.

Possible limitations of our methodology include the fact that we did the HRV analysis within consecutive periods of time, irrespectively of the presence of sleep stage transitions, arousals, awakening, etc. Additionally, we did a post hoc analysis that analyzed the HRV according to different sleep stages only for the first hour of sleep at baseline, R2, and R3 (main outcomes of HRV). Furthermore, our study reported the correlations between SWS and HRV, as well as REM and HRV. This correlation seems to be strengthened more in recovery sleep than in baseline sleep.

Summary

In summary, ANS changes (i.e. HRV and BRS) during recovery sleep, following the accumulation of persistent sleep deficit, reflect cardiovascular recovery processes and appear to continue for at least three nights following resumption of a normal sleep period. The HRV rebound during recovery sleep underscores the importance of sleep for homeostatic regulation of autonomic function.

Funding

This study was supported by National Heart, Lung, and Blood Institute (R01HL106782 to JMM); Harvard Catalyst, Harvard Clinical and Translational Science Center (National Center for Research Resources and National Center for Advancing Translational Sciences, National Institutes of Health Award UL1TR001102).

Conflict of interest statement. None declared.