Does sleep promote adaptation to acute stress: An experimental study

Abstract

Objectives: Evidence of the impact of chronic stress on sleep is abundant, yet experimental sleep studies with a focus on acute stress are scarce and the results are mixed. Our study aimed to fill this gap by experimentally investigating the effects of pre-sleep social stress on sleep dynamics during the subsequent night, as measured with polysomnography (PSG).

Methods: Thirty-four healthy individuals (65% females, M_age = 25.76 years SD = 3.35) underwent a stress-inducing (SC) or neutral control condition (CC) in virtual reality (VR). We used overnight EEG measurements to analyze the basic sleep parameters and power spectral density (PSD) across the sleep cycles, and measured heart rate and its variability (HRV), skin electrodermal activity (EDA), and salivary cortisol to capture physiological arousal during the VR task and the pre-sleep period.

Results: Following acute stress (SC), the amount of slow-wave sleep (SWS) was higher and N2 sleep lower relative to CC, specifically in the first sleep cycle. In SC, PSD was elevated in the beta-low (16-24 Hz) and beta-high (25-35 Hz) frequency ranges during both stages N2 and SWS over the entire night.

Conclusions: Sleep promoted adaptation to acute social stress by a longer duration of SWS in the subsequent sleep period, especially in early sleep. A similar homeostatic effect towards restorative sleep is well-evidenced in animal model stress studies but has not been previously reported in experimental human studies. Whether the high-frequency PSD activity during stages N2 and SWS also serves in the resolution of transient stress, remains open.

Keywords: Experimental study; Polysomnography; Sleep; Stress; Virtual reality.

Fig. 1

Study design. We introduced the participants to the protocol during the preliminary meeting on day 1, assigning their experimental conditions and installing the beat-to-beat and electrodermal activity monitoring devices. Participants spent the subsequent night at home. On day 2, participants spent the daytime freely, arrived at the sleep laboratory at 8 p.m., and completed the VR experiment prior to overnight sleep polysomnography (PSG) measurement in the laboratory. Sleep opportunity was set between 11 p.m. and 7 a.m. One hour after awakening, participants evaluated their sleep quality, and all devices were removed. Cortisol samples and their numbers are indicated by the vial illustrations. Based on averaged times, sample 1 was taken at 08:10 pm. upon arrival at the laboratory. Sample 2 was taken at 08:50 p.m. as a baseline before the VR task. Sample 3 was taken immediately after task completion, and sample 4 was timed exactly 20 min after sample 3 to capture the post-task cortisol peak. Pre-sleep cortisol was assessed with sample 5, taken on average at 11:50 p.m. right before lights off. The mean interval between samples 4 and 5 was 2 h 14 min (SD = 14 min). To assess the cortisol awakening response, samples 6, 7, and 8 were taken on day 3 immediately after waking up at 7 a.m., and 30 and 60 min thereafter.

Fig. 2

Physiological measures during the laboratory evening. Cortisol (A), heart rate (B), and electrodermal activity (C) expressed in relative proportion to the baseline segment, and all figures are adjusted for sex and age. LF = low-frequency, VLF = very-low-frequency, HF = high-frequency, SC = stress condition, CC = control condition, bars refer to 95% confidence interval (CI). **p < .001, *p < .05, +p < .07 linear mixed model condition pairwise comparison and ‘time x condition’ interaction contrasts.

Fig. 3

Standardized linear regression residuals of central delta/beta ratio as a function of time (sleep cycles) during N2 (A) and slow-wave sleep (B), adjusted for age and sex. SC = stress condition, CC = control condition, bars refer to 95% confidence interval (CI). **p < .001, *p < .05 for condition pairwise comparison in linear mixed model, Bonferroni-corrected. Beta = (beta-low + beta-high)/2. Delta (0.5–3 Hz), beta-low (16–24 Hz), and beta-high (25–35 Hz).

Fig. 4

Standardized linear regression residuals of EEG power spectral density (PSD) values averaged over all sleep cycles as function of frequency range, adjusted for age and sex. SC = stress condition, CC = control condition, bars refer to 95% confidence interval (CI). **p < .001, *p < .05 for condition main effect in linear mixed model. Delta (0.5–3 Hz), theta (4–7 Hz), alpha (8–12.5 Hz), sigma (9–16 Hz), beta-low (16–24 Hz), beta-high (25–35 Hz).

Fig. 5

Topographical distribution map of the beta-low and beta-high EEG PSD differences between the conditions in slow-wave sleep. The map displays subtracted values (stress condition minus control condition) calculated with standardized linear regression residuals, adjusted for age and sex. Darker color corresponds to greater difference between the conditions. Reference electrode position was set at FCz. Beta-low (16–24 Hz), beta-high (25–35 Hz).

Fig. 6

Dose-response linear regression between VLF and proportion of SWS (A) and between heart rate and central delta/beta ratio (B), adjusted for age and sex. β = regression coefficient, CI = confidence interval.