3.1. Impact of SWD on Sleep Variables
Measures of sleep quality and quantity for the BSL, SWD, and RCV conditions are summarized in . A repeated-measures ANOVA indicated a treatment effect for sleep stage N3 (% of total sleep time [TST]) and SWA. Post-hoc t-tests revealed a 54% and 37% reduction from BSL to SWD for N3% and SWA, respectively. Other sleep measures, including N1%, N2%, REM%, REM latency (REML), and spindle power (12–15Hz range) were not significantly affected. However, there was also a treatment effect for wake after sleep onset (WASO), arousal index, and sleep efficiency. Post-hoc comparisons showed significantly greater WASO on BSL compared to both SWD and RCV, suggesting a first-night effect. The arousal index was significantly higher on SWD compared to both BSL and RCV, although participants did not report waking up more than usual when questioned about their sleep the following morning. Sleep efficiency was significantly higher on RCV compared to BSL, possibly reflecting a first-night effect on BSL and a sleep rebound on RCV.
Impact of SWD on Sleep, Depressive Symptoms, and Measures of Sleepiness and Alertness (n = 15)
To illustrate the impact of SWD on global SWA across the night compared to the BSL and RCV conditions, all artifact-free NREM sleep epochs for each condition were divided into quintiles, and SWA for each quintile is displayed in . A 3 (condition) × 5 (quintile) repeated-measures ANOVA indicated a main effect for condition (F2,28 = 12.45, p = .0001) and quintile (F4,56 = 22.43, p < .0001), as well as a condition × quintile interaction (F8,112 = 7.19, p < .0001). Post-hoc tests demonstrated that global SWA was lower in the first quintile on SWD compared to both BSL (p < .0001) and RCV (p < .0001), although global SWA in the last quintile did not differ among conditions (p = n.s.).
Figure 1 Global SWA time course across the night for baseline (BSL), slow wave deprivation (SWD), and recovery (RCV) nights. Artifact-free NREM epochs for each night were divided into quintiles. Inset: comparison of SWA dissipation across the entire night for (more ...)
Another measure to evaluate the success of SWD in suppressing SWA involves the dissipation of global SWA across the night. To capture the robust change in SWA that occurs from high sleep pressure to low sleep pressure, which has been linked to distinct changes in neuronal firing patterns that parallel sleep homeostasis (Vyazovskiy et al., 2009
), this study defined SWA dissipation as the SWA of the first quintile minus the SWA of the last quintile. A 1 × 3 ANOVA indicated a significant difference in SWA dissipation among the three study conditions (F2,28
= 15.21, p
< .0001), whereby SWA dissipation was greater on both BSL (p
= .004) and RCV (p
< .0001) compared to SWD (see inset).
3.2. Vigilance Measures
Ruling out confounding effects of alertness or sleepiness that often accompany total sleep deprivation, we found no significant differences on vigilance measures between Pre- and Post-SWD, including the PVT, VAS-Alertness, VAS-Sleepiness, and the SSS. See .
3.3. Decrease in Depressive Symptoms
After one night of SWD, depressed participants showed a significant reduction in depressive symptoms on both the self-rated mIDS-SR (10%) and the HRSD-13 (27%). Individual scores are displayed in . Ensuring that the subset of participants with HRSD-13 scores was representative of the total group of participants, we found similar results for the mIDS-SR in this subset as we found with the entire sample, constituting a 12% reduction in symptoms. When comparing subgroups based on the order of experimental session (whether participants underwent SWD on their 2nd or 4th night during the study), we found no significant differences in the evening pre-sleep scores on the mIDS-SR (16.86 ± 0.46 and 16.25 ± 0.63, respectively). Likewise, there was no significant difference in the overnight decrease of mIDS-SR score (11.03% ± 4.65 and 6.84% ± 2.78, respectively).
Figure 2 Impact of slow wave deprivation (SWD) on the (A) modified Inventory of Depressive Symptomatology, self-rated (mIDS-SR) and (C) Hamilton Rating Scale for Depression (HRSD-13) scores. (B) The amount of REM sleep obtained during the treatment correlates (more ...)
3.4. Predictors of Change in Mood
Because hdEEG was recorded and a treatment effect was observed for global SWA, topographical analyses were conducted to investigate correlations of regional changes in SWA with the overnight change in depressive symptoms. When examining the dissipation of SWA across the night (first – last 20% of total artifact-free NREM sleep epochs) on BSL, the decrease of SWA in a fronto-central cluster of channels () correlated with the decrease in mIDS-SR score. The more SWA was dissipated across the night on baseline, the greater the reduction in depressive symptoms after SWD. As shown in , this correlation was specific for the low SWA frequency range (1–2 Hz) and was inversely related to low spindle power (12–13 Hz).
Figure 3 Topographic correlation of antidepressant response (via mIDS-SR) with (A) dissipation of SWA across the baseline (BSL) night (high sleep pressure – low sleep pressure, expressed as a percentage of SWA averaged across the entire night) and (B) (more ...)
Given that the homeostatic dissipation of SWA on BSL predicted subsequent response to SWD, we sought to establish whether the homeostatic rebound of SWA on RCV also correlated with response. As seen in , greater all-night average SWA on RCV (relative to BSL) in a right frontal cluster of channels was related to greater decreases in symptom severity. Again, this correlation was specific to the low SWA frequency range (< 2Hz) and was inversely related to low spindle power (12–13Hz).
Previous literature (Vogel et al., 1975
) has suggested a role of REM sleep in the antidepressant response to sleep deprivation. Therefore, REM sleep variables were also examined. Despite no significant differences in REM% of total sleep time (), participants with less REM% on SWD showed a greater decrease in mIDS-SR score. See .
To examine the relative impact of baseline SWA dissipation, SWA rebound, and REM%, a multiple regression was conducted with the overnight change in mIDS-SR score as the dependent variable. The results demonstrated that SWA dissipation on BSL [β = 0.30], all-night SWA rebound (BSL – RCV) [β = −0.23], and REM% on SWD [β = −0.32] had similar weights in predicting response.
3.5. Pertinent Negative Findings
Ensuring that the treatment effect for arousal index was not related to the decrease in depressive symptoms, no correlation between the two variables was found (r = 0.262; p > .05). In addition, baseline severity rating on the mIDS-SR did not predict response to SWD (), suggesting that the SWD intervention does not preferentially benefit moderate depression severity (baseline HRSD ≥ 17) compared to more mild levels of depression. Upon questioning during an exit interview at the end of the study, all but one participant reported hearing no tones being played during any of the nights. The one participant who reported hearing tones did not show a greater reduction in depressive symptoms than the group mean, nor did this participant associate a change in mood specific to the SWD night.