-
PDF
- Split View
-
Views
-
Cite
Cite
Charmaine Diep, Suzanne Ftouni, Jessica E Manousakis, Christian L Nicholas, Sean P A Drummond, Clare Anderson, Acoustic slow wave sleep enhancement via a novel, automated device improves executive function in middle-aged men, Sleep, Volume 43, Issue 1, January 2020, zsz197, https://doi.org/10.1093/sleep/zsz197
Close -
Share
Abstract
As slow-wave activity (SWA) is critical for cognition, SWA-enhancing technologies provide an exciting opportunity to improve cognitive function. We focus on improving cognitive function beyond sleep-dependent memory consolidation, using an automated device, and in middle-aged adults, who have depleted SWA yet a critical need for maximal cognitive capacity in work environments.
Twenty-four healthy adult males aged 35–48 years participated in a randomized, double-blind, cross-over study. Participants wore an automated acoustic stimulation device that monitored real-time sleep EEG. Following an adaptation night, participants were exposed to either acoustic tones delivered on the up phase of the slow-wave (STIM) or inaudible “tones” during equivalent periods of stimulation (SHAM). An executive function test battery was administered after the experimental night.
STIM resulted in an increase in delta (0.5–4 Hz) activity across the full-night spectra, with enhancement being maximal at 1 Hz. SWA was higher for STIM relative to SHAM. Although no group differences were observed in any cognitive outcomes, due to large individual differences in SWA enhancement, higher SWA responders showed significantly improved verbal fluency and working memory compared with nonresponders. Significant positive associations were found between SWA enhancement and improvement in these executive function outcomes.
Our study suggests that (1) an automated acoustic device enhances SWA; (2) SWA enhancement improves executive function; (3) SWA enhancement in middle-aged men may be an important therapeutic target for enhancing cognitive function; and (4) there is a need to examine interindividual responses to acoustic stimulation and its effect on subsequent cognitive function.
This study has been registered with the Australian New Zealand Clinical Trials Registry. “The efficacy of acoustic tones in slow-wave sleep enhancement and cognitive function in healthy adult males”. https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=371548&isReview=true
ACTRN12617000399392.
Slow-wave sleep is critical for cognition. Middle age represents a unique target for slow-wave activity (SWA) enhancement due to depleted SWA and a need for optimal cognitive performance in the workplace. Our study describes a novel, automated acoustic stimulation device that enhances SWA and subsequently enhances executive function in middle-aged men. We provide evidence of a potential benefit of an automated device that can be readily deployed in at-home settings to enhance SWA and enhance cognitive function in the wider community.
Introduction
Nonrapid eye movement (NREM) slow-wave sleep (SWS or N3) is homeostatically regulated and tightly associated with the duration and intensity of the waking day [1–4]. The electrophysiological signature of SWS is slow-wave activity (SWA), typically referred to as EEG power between 0.5 and 4.5 Hz, and its magnitude and dynamics are considered the primary marker of sleep homeostasis [1]. Beyond simply reflecting homeostatic sleep pressure, SWA serves a fundamental purpose to human function and behavior as it is critical for optimal cognitive functioning [5]. The link between SWA and cognitive function has prompted the examination of new strategies targeting the augmentation of SWA to maximize cognitive function, particularly in later life [6, 7] where SWA is severely depleted.
The link between SWA and cognitive performance has been well documented in older adults [5, 8], particularly their parallel inter-connected decline [9]. SWS however declines steadily from early adulthood [10, 11]: On average, SWS occupies 18.9% of the sleep period in young adults (16–25 years), and declines 2–3% per decade to occupy only 3.4% of the sleep period by middle age (35–50 years). Despite the marked reduction in SWA in mid-life, most studies of SWA and cognitive function have focused on either younger or older adults, where SWA is either maximal or depleted. Moreover, mid-life is a period where optimal cognitive capacity is required with respect to employment and productivity. For instance, more middle-aged adults are active in the workforce compared to younger (~30% more) and older (~60% more) adults [12], and middle-aged adults are at peak earning potential, relative to both younger and older adults [13]. Given the gradual depletion of SWA in this age group, easily employed techniques to enhance SWA and improve cognitive capability have important social and economic implications for middle-aged adults.
A number of approaches to SWA enhancement have been applied successfully, e.g., transcranial magnetic stimulation [14], transcranial direct stimulation [15], and pharmaceutical methods [16]. Although pharmacological or slow-oscillatory electrical stimulation approaches are effective at enhancing SWA, they present several limitations, including side effects such as sleep inertia, or are not feasible for application within the home. A more recent approach to SWA is acoustic stimulation. This method of enhancement provides a non-invasive tool that is cost-effective with no known side effects and with recent novel technological advancements can now be easily applied in the home setting.
Acoustic stimulation of SWA involves pulses of auditory tones specifically targeted to the upstate of the intrinsically generated ‘natural’ slow-wave oscillation which then entrain the endogenous slow-wave and increase its amplitude [17]. This has been shown in young [17–20] and older [7] adults. Importantly, acoustic enhancement of SWA has been associated with improved cognitive outcomes, particularly declarative memory [7, 17, 19], largely due to the important role SWA plays in the consolidation of declarative memories [5, 21, 22]. The benefit of SWA has also been linked to visuo-motor learning [23], perceptual learning [24], and executive function [8]; for which SWA-rich prefrontal areas [25] are essential for optimum function [26]. The extent to which SWA augmentation via acoustic stimulation benefits such cognitive functions beyond overnight memory consolidation is unknown.
Our study aim was to examine the effectiveness of an automated, self-regulating acoustic SWA stimulation device (SmartSleep, Philips Healthcare) in middle-aged men, and to evaluate any improvement in overnight memory consolidation and executive function. To achieve this, men between 35 and 50 years of age wore the device for two consecutive nights, repeated on two separate occasions 1 week apart: the first night acting as a baseline night and the second an experimental night where the device operated in either STIM (acoustic tones delivered) or SHAM (device operational but no audible tones delivered) mode. Day time cognitive functioning measures were evaluated following the experimental nights.
Material and Methods
Participant Screening and Recruitment
Twenty-four healthy male participants (aged 35–48 years, mean ± SD 39.92 ± 4.15 years) were recruited from the general population. Exclusion criteria included known history of any psychiatric or mood disorders including immediate family members, past or current presence of sleep disorders, current hypnotic or psychoactive drug use, hearing impairments, current smoker or use of nicotine therapies, a history of loss of consciousness greater than 15 minutes and transmeridian travel or shift work within the past 3 months of participation. Naps, alcohol, and caffeine consumption were limited to 2 naps per week, 14 standard drinks per week and 300 mg per day, respectively. Few participants napped during the study (4/24), and all naps occurred at least 3 days prior to laboratory admit and was consistent between conditions. Participants were within normal ranges for self-reported sleep quality, daytime sleepiness and mood (<5 Pittsburgh Sleep Quality Index [27]; <10 Epworth Sleepiness Scale [28], <5 Patient Health Questionnaire-9 [29], and Depression Anxiety and Stress Scale-21 (Depression: <6, Anxiety: <5, Stress: <10 [30]). As a check on hearing, all participants reported hearing the tones from the device at consent. All participants gave written informed consent and were compensated for their time. The study was approved by the Monash University Human Research Ethics Committee #CF15/671 – 2015000308.
Experimental Design
The full study protocol is summarized in Figure 1. We used a randomized, double-blind, sham-controlled, cross-over design whereby participants were exposed to a two-night protocol on two separate occasions separated by at least 5 days wash-out between visits. Participants not admitted on the same day of the week still had weekday admits (as opposed to weekend nights) as weekend/weekday sleep may differ [31]. As SWA is homeostatically regulated and tightly coupled to the length of the waking day [32], strict adherence to sleep/wake schedule based on their habitual sleep/wake cycle was required for 1 week prior to admittance into the laboratory. Participants were required to have a sleep efficiency of at least 80% and to comply with their sleep/wake schedule. During this period, they abstained from caffeine and alcohol for 48 hours prior to and during each laboratory visit. Upon admission to the study, participants completed a drug and alcohol screen, and sleep adherence was monitored through monitored through Actiwatch Spectrums (Philips Respironics), sleep diaries and time-stamped call-ins. Full polysomnography including respiratory monitoring was conducted on Night 1 to screen for sleep apnea (Apnea–Hypopnea Index < 10) and periodic leg movements (PLMs) (PLMI with arousals < 5) with an 8-hour sleep opportunity starting at habitual sleep time.
Depiction of laboratory protocol. On Day 1 of the protocol, participants completed Attention and Vigilance (AV) test batteries from 1.5 h post-wake and every 2 h thereafter. Karolinska Sleepiness Scale (KSS) was administered hourly. Participants undertook the encoding part of the paired associate task 2.5 h prior to habitual sleep time. On Day 2, attention and vigilance test batteries occurred at the same time. The paired associate task recall was administered at 1 h post-wake, and neurocognitive testing batteries occurred at 2 and 4 h post-wake. Participants were discharged approximately 10 h post-wake on Day 2.
Depiction of laboratory protocol. On Day 1 of the protocol, participants completed Attention and Vigilance (AV) test batteries from 1.5 h post-wake and every 2 h thereafter. Karolinska Sleepiness Scale (KSS) was administered hourly. Participants undertook the encoding part of the paired associate task 2.5 h prior to habitual sleep time. On Day 2, attention and vigilance test batteries occurred at the same time. The paired associate task recall was administered at 1 h post-wake, and neurocognitive testing batteries occurred at 2 and 4 h post-wake. Participants were discharged approximately 10 h post-wake on Day 2.
Acoustic Stimulation
The device monitored EEG Fpz, referenced to Fp1 and M1. Acoustic volume was between 20 and 65 dB and was individualized such that the device automatically calibrated the auditory threshold on the baseline night to ensure the tone was sufficient to provide augmentation while not inducing an arousal. In the active condition (STIM), the device would administer an initial tone phase-locked to the up phase of the slow wave followed by tones at a frequency of 1 Hz, as described previously [33]. This continued until the end of the N3 period or if an arousal was detected and was repeated for each bout of SWS (N3). In the inactive condition (SHAM), the device monitored sleep without administering audible tones.
Polysomnography Recording
Electroencephalography (EEG), electrooculography, electromyography, and electrocardiography were recorded with Profusion polysomnography (PSG) 4 (Compumedics, Melbourne, Australia) and Grass gold-cup electrodes from 18 channels using the international 10–20 system (Fp1, Fp2, F3, F4, C3, C4, P3, P4, PO3, PO4, O1, O2, Fpz, Fz, Cz, Cpz, Pz, Oz). Data were sampled at 512 Hz; impedances were maintained at ≤5 Ω. On both baseline nights, thoracic and abdominal respiratory effort, airflow, finger pulse oximetry, and bilateral leg movements were monitored for diagnostic purposes.
Cognitive Tasks
Sleep-dependent memory was assessed with an episodic word-pair task with an encoding and criterion phase administered approximately 2 hours prior to bedtime, a short-delay recognition trial administered approximately 1 hour prior to bed time, and a long-delay recognition trial administered 1 hour post-wake. Other tasks were run in two test batteries to reduce any time on task fatigue [34]. The Tower of London and Verbal Fluency were administered in the first test battery (2 hours post-wake) and the Go No Go and N-Back (4 hours post-wake) were administered in the second battery. All tasks had alternate versions and order of administration was counterbalanced.
Paired Associate Learning Memory Test
The Paired Associate Learning measure of declarative memory uses 120 word-nonsense word pairs instead of semantically linked word pairs to encourage hippocampal-dependent learning [5]. The memory score for the short- and long-delay recall tasks was calculated as [number of words correctly identified − lure words − foil words]. The change score is calculated as [morning memory score − evening memory score].
Tower of London
The Tower of London measures nonverbal planning as described in Shallice [35]. The computerized version involved moving a set of blocks from a standard pattern to a target pattern in a limited number of moves. Participants completed six problems ranging in difficulty as per ref. [36] with the primary outcome measure being planning time.
Verbal Fluency
Participants had 60 seconds to verbally respond with words to a given prompt. Phonetic fluency was measured across three trials using letters as prompts (e.g. FAS and BHR), and semantic fluency was measured across two trials (animals and boys names; musical instruments and girls names), with number of correct words summed across trials. A category switching trial was administered, whereby participants had to switch between two categories (e.g. fruit and furniture), and number of accurate switches was calculated [37].
Go No Go
The Go No Go task is a measure of inhibition whereby participants must respond to three shapes (e.g. big triangle, small triangle, big star; n = 124) and withhold response to a fourth shape (small star; n = 57), which resembles the others in form or size [38, 39]. The primary outcome measure is d’, which is a ratio of correct responses to false (failure to inhibit) responses.
N-back
The N-back is a test of working memory, where participants are shown a randomized sequence of characters of the alphabet and must respond to characters that match those shown either immediate prior (1-back) or 2 positions back (2-back) [40].
Attention and Vigilance Battery
Each Attention and Vigilance battery consisted of a 10-minute psychomotor vigilance test, a Karolinska drowsiness test (3 minutes eyes open, 2 minutes eyes closed), and a Karolinska Sleepiness Scale at the beginning and end of each battery. This was not a primary outcome of the study but was used to check consistency between test sessions given the impact of sleep on cognitive alertness [41].
Sleep and Spectral Analysis
Sleep staging, sleep onset, and arousals were scored visually according to standard American Academy of Sleep Medicine criteria [42], channels were referenced to contralateral mastoid (e.g. C3-M1, C4-M1), and data were filtered between 0.3 and 30 Hz, with a 50 Hz Notch filter. Scorers were blinded to the experimental condition. Minutes of total sleep time, N1, N2, N3, rapid eye movement (REM) sleep, wake after sleep onset (WASO), and number of arousals were calculated from the scored data. Sleep cycles were determined according to previously used criteria [43]. Two participants had cycles 1 and 2 combined due to short (<15 minutes) first cycles (this was observed for both STIM and SHAM conditions; thus, data were treated equally for both conditions). Spectral analyses were conducted within Curry 7 software (Compumedics NeuroScan, Melbourne, Australia). Data were visually cleaned to remove artifact in 5-second windows. EEG data were digitally re-referenced to linked mastoids and then bandpass filtered with a zero-phase shift Hann filter from 0.3 to 30 Hz with 0.6 and 8 Hz slopes, respectively. Power spectra was calculated in nonoverlapping 5-second epochs, resulting in a 0.125 Hz resolution. Total absolute power was determined by averaging raw power across the night within each frequency bin and summed across frequency bands (0.5–4 Hz, etc.). As night-to-night variability in absolute power can have a large impact on EEG power [44, 45], we calculated relative power by expressing SWA as a function of total power within the REM bandwidth. Although other EEG studies may calculate relative power as a function of the total power spectra, this was problematic due to the power spectra of one condition (STIM) being systematically altered by the study condition. Expressing data relative to the full power spectra therefore would have canceled out the effect of the acoustic stimulation, i.e., if whole night data were used as the denominator for normalization, enhanced SWA would be both in the numerator and the denominator, thereby canceling itself out. Therefore, expressing EEG data relative to total power spectra in the absence of stimulation (i.e. during REM sleep) was utilized (we did not use N1 as this was associated with stage transitions, and N2 had, on occasion, acoustic tones delivered). For each frequency bin, relative EEG was calculated as absolute NREM δ power − x̅ REM δ/SD REM δ. NREM was defined as NREM 2 + 3. As a check on the reliability of using REM for relative EEG, we replicated analyses with absolute data, checked REM power spectra across both conditions, and examined night-to-night stability in absolute delta power in NREM 2 + 3 relative to REM (i.e. to ensure one night does not involve an increase in NREM SWA which is absent in REM SWA), by assessing the ratio of NREM/REM delta power for baseline week 1 and baseline week 2.
Slow-wave energy (SWE) was calculated as SWA*#minutes N2+N3 for each cycle, which was then summed for each successive sleep cycle. As SWA reduces with each concurrent N3 cycle (due to the homeostatic reduction of SWA), examining SWA in full-night spectra can be artificially lowered for those exhibiting additional N3 sleep cycles. As this may be of benefit for those exhibiting SWA augmentation, we calculated SWE. SWE described the cumulative change in SWA (0.5–4 Hz) across the first four sleep cycles, while taking into account total number of (N2 + N3) epochs. Primary analyses are derived from C3 (except for n = 1 where C4 was substituted for both conditions due to loss of C3) as per previous studies of acoustic stimulation [20] and as k-complexes are maximal over fronto-central sites [46], and frontal sites were associated with more noise due to the placement of the acoustic stimulation headband.
Statistical Analysis
IBM SPSS Statistics V23 and GraphPad Prism V7.0 were used to run statistical analyses. Normality assumptions were checked using SPSS. Cycles were analyzed via a linear mixed model, with cycle as a fixed factor. Paired t-tests were run comparing SHAM and STIM if normality assumptions were met. Nonparametric data were analyzed using Mann–Whitney U or Wilcoxon matched-pair tests. Pearson’s correlations were used to compare the relationship between SWE percent change and cognitive test percent change. Univariate outliers within cognitive results were defined as mean ± (2.5 × SD) and were corrected as next most extreme +1 [47]. Multivariate outliers were detected with Cook’s distance and further examined to determine whether they were a valid case (i.e. not caused by artifact or study violations). To correct for multiple comparisons, we applied Benjamini–Hochberg’s cut-offs for false discovery rate [48]. Unless stated otherwise, all results presented are mean ± SEM. Results under p < .05, two-tailed were considered significant.
Data Retention
One participant (4%) was not included in baseline sleep parameters due to missing data. For device operations (e.g. number of tones, Figure 2), there were missing data for two participants (total 8% data loss). One participant was excluded from analyses due to external sleep disturbances. For cognitive testing, participants were removed due to missing data or technical errors and subsequently when checked for outliers (according to Kolmogorov–Smirnov or Cooks distance). For verbal fluency, the phonetic fluency was administered incorrectly on two occasions (out of 48, 4.2% data loss) resulting in the loss of two (8.4%) (12.5%) participants. For the remaining cognitive tests, where data were statistical outliers only, results are presented with and without outliers for clarity.
Relationship between device activity and sleep parameters. (A) Significant correlations between device activation and SWS were observed for STIM and (B) SHAM nights. A reduced number of arousals were associated with increased device activation for (C) STIM, but not (D) SHAM. Error is 95% confidence intervals of the regression line.
Relationship between device activity and sleep parameters. (A) Significant correlations between device activation and SWS were observed for STIM and (B) SHAM nights. A reduced number of arousals were associated with increased device activation for (C) STIM, but not (D) SHAM. Error is 95% confidence intervals of the regression line.
Results
Sleep Architecture Prior to Acoustic Stimulation
Due to the strong homeostatic regulation of SWS, we first confirmed there were no significant differences in actigraphically determined sleep during the week of at-home sleep preceding admission to the laboratory protocol (p > .2), nor any differences in PSG-defined sleep outcomes between STIM and SHAM for the baseline night (p > .2). See Table 1. In addition, there were no significant differences in relative whole night spectra when comparing the baseline nights for each condition (SHAM: 520.7 ± 59.6 µV2 vs. STIM: 479.3 ± 65.2 µV2, p = .24). See Supplementary Figure S1.
Sleep outcomes for the one week prior to the study, and the baseline night to check consistency between conditions.
| Parameter | STIM | SHAM | P | |
|---|---|---|---|---|
| One-week | Sleep onset | 22.7 ± 0.2 | 22.7 ± 0.2 | .3 |
| pre-study | Sleep offset | 6.3 ± 0.1 | 6.4 ± 0.1 | .42 |
| (Actigraphy) | Sleep efficiency | 82.3 ± 1.7 | 83.5 ± 1.5 | .92 |
| Sleep latency | 9.8 ± 1.6 | 9.4 ± 1.2 | .54 | |
| WASO | 56.0 ± 6.8 | 56.1 ± 7.9 | .9 | |
| TST | 6.3 ± 0.1 | 6.5 ± 0.1 | .8 | |
| Baseline | N1 | 22.4 ± 3.3 | 21.04 ± 2.5 | .74 |
| night PSG | N2 | 218.8 ± 6.8 | 221.89 ± 10 | .76 |
| Parameters | N3 | 82.2 ± 7.6 | 80.87 ± 7 | .86 |
| REM | 90.7 ± 4.4 | 91.57 ± 6.7 | .9 | |
| Sleep latency | 8.72 ± 1.1 | 8.98 ± 1.7 | .88 | |
| Arousal Index | 11.92 ± 1.2 | 11.56 ± 1.1 | .74 | |
| WASO | 52.93 ± 7. 9 | 51.74 ± 6.7 | .82 | |
| TST (min) | 6.9 ± 0.1 | 6.92 ± 0.1 | .84 |
| Parameter | STIM | SHAM | P | |
|---|---|---|---|---|
| One-week | Sleep onset | 22.7 ± 0.2 | 22.7 ± 0.2 | .3 |
| pre-study | Sleep offset | 6.3 ± 0.1 | 6.4 ± 0.1 | .42 |
| (Actigraphy) | Sleep efficiency | 82.3 ± 1.7 | 83.5 ± 1.5 | .92 |
| Sleep latency | 9.8 ± 1.6 | 9.4 ± 1.2 | .54 | |
| WASO | 56.0 ± 6.8 | 56.1 ± 7.9 | .9 | |
| TST | 6.3 ± 0.1 | 6.5 ± 0.1 | .8 | |
| Baseline | N1 | 22.4 ± 3.3 | 21.04 ± 2.5 | .74 |
| night PSG | N2 | 218.8 ± 6.8 | 221.89 ± 10 | .76 |
| Parameters | N3 | 82.2 ± 7.6 | 80.87 ± 7 | .86 |
| REM | 90.7 ± 4.4 | 91.57 ± 6.7 | .9 | |
| Sleep latency | 8.72 ± 1.1 | 8.98 ± 1.7 | .88 | |
| Arousal Index | 11.92 ± 1.2 | 11.56 ± 1.1 | .74 | |
| WASO | 52.93 ± 7. 9 | 51.74 ± 6.7 | .82 | |
| TST (min) | 6.9 ± 0.1 | 6.92 ± 0.1 | .84 |
All results in decimal numbers unless stated otherwise. WASO, wake after sleep onset; TST, total sleep time; REM, rapid eye movement.
Sleep outcomes for the one week prior to the study, and the baseline night to check consistency between conditions.
| Parameter | STIM | SHAM | P | |
|---|---|---|---|---|
| One-week | Sleep onset | 22.7 ± 0.2 | 22.7 ± 0.2 | .3 |
| pre-study | Sleep offset | 6.3 ± 0.1 | 6.4 ± 0.1 | .42 |
| (Actigraphy) | Sleep efficiency | 82.3 ± 1.7 | 83.5 ± 1.5 | .92 |
| Sleep latency | 9.8 ± 1.6 | 9.4 ± 1.2 | .54 | |
| WASO | 56.0 ± 6.8 | 56.1 ± 7.9 | .9 | |
| TST | 6.3 ± 0.1 | 6.5 ± 0.1 | .8 | |
| Baseline | N1 | 22.4 ± 3.3 | 21.04 ± 2.5 | .74 |
| night PSG | N2 | 218.8 ± 6.8 | 221.89 ± 10 | .76 |
| Parameters | N3 | 82.2 ± 7.6 | 80.87 ± 7 | .86 |
| REM | 90.7 ± 4.4 | 91.57 ± 6.7 | .9 | |
| Sleep latency | 8.72 ± 1.1 | 8.98 ± 1.7 | .88 | |
| Arousal Index | 11.92 ± 1.2 | 11.56 ± 1.1 | .74 | |
| WASO | 52.93 ± 7. 9 | 51.74 ± 6.7 | .82 | |
| TST (min) | 6.9 ± 0.1 | 6.92 ± 0.1 | .84 |
| Parameter | STIM | SHAM | P | |
|---|---|---|---|---|
| One-week | Sleep onset | 22.7 ± 0.2 | 22.7 ± 0.2 | .3 |
| pre-study | Sleep offset | 6.3 ± 0.1 | 6.4 ± 0.1 | .42 |
| (Actigraphy) | Sleep efficiency | 82.3 ± 1.7 | 83.5 ± 1.5 | .92 |
| Sleep latency | 9.8 ± 1.6 | 9.4 ± 1.2 | .54 | |
| WASO | 56.0 ± 6.8 | 56.1 ± 7.9 | .9 | |
| TST | 6.3 ± 0.1 | 6.5 ± 0.1 | .8 | |
| Baseline | N1 | 22.4 ± 3.3 | 21.04 ± 2.5 | .74 |
| night PSG | N2 | 218.8 ± 6.8 | 221.89 ± 10 | .76 |
| Parameters | N3 | 82.2 ± 7.6 | 80.87 ± 7 | .86 |
| REM | 90.7 ± 4.4 | 91.57 ± 6.7 | .9 | |
| Sleep latency | 8.72 ± 1.1 | 8.98 ± 1.7 | .88 | |
| Arousal Index | 11.92 ± 1.2 | 11.56 ± 1.1 | .74 | |
| WASO | 52.93 ± 7. 9 | 51.74 ± 6.7 | .82 | |
| TST (min) | 6.9 ± 0.1 | 6.92 ± 0.1 | .84 |
All results in decimal numbers unless stated otherwise. WASO, wake after sleep onset; TST, total sleep time; REM, rapid eye movement.
Acoustic Stimulation and Sleep Parameters—STIM Versus SHAM
We then examined the difference between STIM and SHAM device activation by examining the number of delivered tones and number of minutes of activation during each experimental night. No differences were observed in the number of tones delivered during STIM (2489.9 ± 408.5) or SHAM (2456.4 ± 373.4, p = .38), with a 0.56-minute difference in device activation between the two nights (STIM: 40.9 ± 6.2 minutes and SHAM: 41.5 ± 6.8 minutes), noting that the device was active, yet inaudible for SHAM. Strong correlations between STIM and SHAM for device activation (r = .9, p < .0001) were observed. As expected, given tone administration is dependent on the detection of the slow wave, device activation (number of minutes) was positively correlated with minutes of SWS during both STIM (r = .74, p < .0001) and SHAM nights (r = .56, p < .01). See Figure 2A and B, respectively. Device activation (number of minutes) was associated with a decrease in arousals during STIM (r = −.47, p < .05), but not SHAM (r = −.32, p = .07) (see Table 2). See Figure 2C and D, respectively. During STIM, increased device activation was also associated with a decrease in N1 and N2 sleep (r = −.49 and −.44, respectively, p < .05). No other associations were found between device activation and other sleep parameters, for STIM or SHAM mode (see Table 2).
Pearson’s correlations between device activation and PSG-sleep parameters
| STIM | SHAM | |||
|---|---|---|---|---|
| Parameter | r | P | r | P |
| N1 | −.49 | .019 | −.49 | .08 |
| N2 | −.44 | .035 | −.44 | .27 |
| N3 | .74 | <.0001 | .56 | .003 |
| REM | −.22 | .305 | −.29 | .2 |
| Arousal Index | −.47 | .027 | −.32 | .07 |
| WASO | −.098 | .647 | .16 | .46 |
| TST | −.015 | .946 | −.18 | .23 |
| STIM | SHAM | |||
|---|---|---|---|---|
| Parameter | r | P | r | P |
| N1 | −.49 | .019 | −.49 | .08 |
| N2 | −.44 | .035 | −.44 | .27 |
| N3 | .74 | <.0001 | .56 | .003 |
| REM | −.22 | .305 | −.29 | .2 |
| Arousal Index | −.47 | .027 | −.32 | .07 |
| WASO | −.098 | .647 | .16 | .46 |
| TST | −.015 | .946 | −.18 | .23 |
WASO, wake after sleep onset; TST, total sleep time; REM, rapid eye movement.
Pearson’s correlations between device activation and PSG-sleep parameters
| STIM | SHAM | |||
|---|---|---|---|---|
| Parameter | r | P | r | P |
| N1 | −.49 | .019 | −.49 | .08 |
| N2 | −.44 | .035 | −.44 | .27 |
| N3 | .74 | <.0001 | .56 | .003 |
| REM | −.22 | .305 | −.29 | .2 |
| Arousal Index | −.47 | .027 | −.32 | .07 |
| WASO | −.098 | .647 | .16 | .46 |
| TST | −.015 | .946 | −.18 | .23 |
| STIM | SHAM | |||
|---|---|---|---|---|
| Parameter | r | P | r | P |
| N1 | −.49 | .019 | −.49 | .08 |
| N2 | −.44 | .035 | −.44 | .27 |
| N3 | .74 | <.0001 | .56 | .003 |
| REM | −.22 | .305 | −.29 | .2 |
| Arousal Index | −.47 | .027 | −.32 | .07 |
| WASO | −.098 | .647 | .16 | .46 |
| TST | −.015 | .946 | −.18 | .23 |
WASO, wake after sleep onset; TST, total sleep time; REM, rapid eye movement.
Acoustic Enhancement of SWA
For the experimental night, we observed no difference in PSG-derived sleep parameters between STIM and SHAM conditions (see Table 3). We found no increase in absolute SWA in STIM relative to SHAM (p = .15) for full-night spectra, although large increases were found for cycle 4 for both delta power (0.5–4 Hz—SHAM: 185.01 µV2 vs. STIM: 206.04 µV2, p < .05, d = 0.45) and low-frequency delta (<1 Hz—SHAM: 143.55 µV2 vs. STIM: 147.11 µV2, p < .05, d = 0.51). No differences were observed in cycles 1–3 (p > .2), although cycle 2 had a moderate effect size (d = 0.25). Given that analysis of absolute data is not ideal for this study design (i.e. conditions on laboratory visits and large individual difference [44]), we focused our analyses on relative data [which was positively correlated to the absolute data for SHAM (r = .5, p = .01) and STIM (r = .6, p = .003)]. To demonstrate comparable changes in the absolute data however, we have replicated all analyses with absolute data as shown in Supplementary Figures S3–S5.
Sleep parameters for experimental nights
| Parameter | STIM | SHAM | P |
|---|---|---|---|
| N1 | 13.33 ± 1.9 | 15.25 ± 1.5 | .25 |
| N2 | 225.21 ± 5.7 | 221.83 ± 7.7 | .69 |
| N3 | 59.67 ± 6.6 | 59.13 ± 7.1 | .88 |
| REM | 112.63 ± 5.8 | 112.90 ± 5.4 | .96 |
| Sleep latency | 12.94 ± 1.7 | 12.50 ± 1.5 | .77 |
| Arousal Index | 12.52 ± 6.3 | 12.56 ± 2 | .96 |
| WASO | 49.71 ± 4.9 | 53.71 ± 6 | .53 |
| TST | 410.83 ±1.3 | 413.25 ± 32.9 | .77 |
| Parameter | STIM | SHAM | P |
|---|---|---|---|
| N1 | 13.33 ± 1.9 | 15.25 ± 1.5 | .25 |
| N2 | 225.21 ± 5.7 | 221.83 ± 7.7 | .69 |
| N3 | 59.67 ± 6.6 | 59.13 ± 7.1 | .88 |
| REM | 112.63 ± 5.8 | 112.90 ± 5.4 | .96 |
| Sleep latency | 12.94 ± 1.7 | 12.50 ± 1.5 | .77 |
| Arousal Index | 12.52 ± 6.3 | 12.56 ± 2 | .96 |
| WASO | 49.71 ± 4.9 | 53.71 ± 6 | .53 |
| TST | 410.83 ±1.3 | 413.25 ± 32.9 | .77 |
All results in minutes unless stated otherwise. WASO, wake after sleep onset; TST, total sleep time; REM, rapid eye movement.
Sleep parameters for experimental nights
| Parameter | STIM | SHAM | P |
|---|---|---|---|
| N1 | 13.33 ± 1.9 | 15.25 ± 1.5 | .25 |
| N2 | 225.21 ± 5.7 | 221.83 ± 7.7 | .69 |
| N3 | 59.67 ± 6.6 | 59.13 ± 7.1 | .88 |
| REM | 112.63 ± 5.8 | 112.90 ± 5.4 | .96 |
| Sleep latency | 12.94 ± 1.7 | 12.50 ± 1.5 | .77 |
| Arousal Index | 12.52 ± 6.3 | 12.56 ± 2 | .96 |
| WASO | 49.71 ± 4.9 | 53.71 ± 6 | .53 |
| TST | 410.83 ±1.3 | 413.25 ± 32.9 | .77 |
| Parameter | STIM | SHAM | P |
|---|---|---|---|
| N1 | 13.33 ± 1.9 | 15.25 ± 1.5 | .25 |
| N2 | 225.21 ± 5.7 | 221.83 ± 7.7 | .69 |
| N3 | 59.67 ± 6.6 | 59.13 ± 7.1 | .88 |
| REM | 112.63 ± 5.8 | 112.90 ± 5.4 | .96 |
| Sleep latency | 12.94 ± 1.7 | 12.50 ± 1.5 | .77 |
| Arousal Index | 12.52 ± 6.3 | 12.56 ± 2 | .96 |
| WASO | 49.71 ± 4.9 | 53.71 ± 6 | .53 |
| TST | 410.83 ±1.3 | 413.25 ± 32.9 | .77 |
All results in minutes unless stated otherwise. WASO, wake after sleep onset; TST, total sleep time; REM, rapid eye movement.
As EEG data were expressed relative to REM power spectra (as a clean, nonenhanced segment of EEG), we first ensured that changes in REM SWA were not driving any observed differences. First, we checked that there were no differences in power spectra for STIM versus SHAM for either the total REM power spectra (SHAM: 82.75 ± 6.38 µV2 vs. STIM: 80.16 ± 6.24 µV2, p = .31) or REM delta power (SHAM: 58.58 ± 6.02 µV2 vs. 54.53 ± 4.75 µV2, p = .23). Second, we demonstrated that the ratio of absolute delta power in NREM and REM sleep on the baseline nights was stable (p > .15), suggesting that the differences in relative SWA for STIM are not simply driven by an instability in NREM SWA and REM SWA week to week. As seen in Figure 3A, for whole night spectra we observed significantly increased power within the delta bandwidth (0.5–4 Hz) for STIM compared to SHAM (t(22) = 2.65, p = .02, d = 0.65). There was also decreased power in the alpha (8-12 Hz) (t(22) = 2.57, p = .02, d = 0.84) and slow-sleep spindle (low sigma: 12–14 Hz) bandwidths (t(23) = 2.71, p = .02, d = 0.87). The magnitude of percentage change was high for delta, relative to other spectral bandwidths (see Figure 3B), corresponding to an 11.6% increase in delta power in STIM relative to the SHAM night. Within the delta bandwidth (Figure 3C), STIM enhanced SWA for the majority of 0.5 Hz bins (particularly <2 Hz) with small-to-moderate effect sizes (d = 0.15–0.52). Peak enhancement during the STIM night was found at 1 Hz (t(23) = 2.7, p = .04, d = 0.62), and decreased with each 0.5 Hz bin thereafter. Absolute data are shown in Supplementary Figure S3A–C. There were no significant changes observed for parietal regions (SHAM: 242.1 ± 109.5 µV2 vs. STIM: 259.9 ± 66.9 µV2, t(22) = 0.69, p > .1), and an increase observed in the frontal region (SHAM: 352.5 ± 28.9 µV2 vs. STIM: 378.7 ± 29.9 µV2, t(22) = 2.3, p = .02).
Acoustic enhancement of SWA during STIM relative to SHAM experimental nights. (A) Differences in relative power in full-night spectra for STIM night relative to SHAM, in 0.5 Hz bins normalized to delta power during all-night REM for all power spectra and (B) for each predetermined EEG spectral bandwidth including delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), slow spindles (SS, 12–14 Hz), fast spindles (FS, 14–16 Hz), and beta (16–30 Hz). (C) Enhancement of SWA in the delta range was greatest at 1.0 Hz, relative to other delta frequencies.
Acoustic enhancement of SWA during STIM relative to SHAM experimental nights. (A) Differences in relative power in full-night spectra for STIM night relative to SHAM, in 0.5 Hz bins normalized to delta power during all-night REM for all power spectra and (B) for each predetermined EEG spectral bandwidth including delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), slow spindles (SS, 12–14 Hz), fast spindles (FS, 14–16 Hz), and beta (16–30 Hz). (C) Enhancement of SWA in the delta range was greatest at 1.0 Hz, relative to other delta frequencies.
To examine changes across the night, we examined SWA in the first four sleep cycles. SWA was higher during STIM in cycle 2 (STIM: 623.05 ± 79.01 µV2, SHAM: 512.51 ± 55.83 µV2, Wilcoxon signed ranks test, z = −2.68, p = .025, d = 0.71) and cycle 4 (STIM: 270.74 ± 31.09 µV2, SHAM: 208.62 ± 21.64 µV2, t(22) = 3.46, p = .01, d = 0.84), whereas no enhancements were observed in cycle 1 (p > .1) or cycle 3 (p > .1) (see above for absolute data). Only 33% of participants in SHAM experienced a fourth N3 cycle, whereas almost twice as many of these same individuals (58%) had a fourth N3 cycle following STIM. We therefore examined SWE to show cumulative changes across the night. Here, SWE was higher in STIM relative to SHAM (t(22) = 2.39, p < .5, d = 0.54), with a moderate effect size (Figure 4). There were no differences in the distribution of sleep stages within each sleep cycle for STIM compared with SHAM, suggesting that the increase in SWE was not due to altered sleep staging (see Supplementary Figure S4).
Change in slow-wave energy for STIM versus SHAM. (A) STIM increased slow-wave energy (SWA*#mins of N2+N3) across the full-night spectra and from the second sleep cycle compared with (B) SHAM. *p < .05. Error bars indicate SEM.
Change in slow-wave energy for STIM versus SHAM. (A) STIM increased slow-wave energy (SWA*#mins of N2+N3) across the full-night spectra and from the second sleep cycle compared with (B) SHAM. *p < .05. Error bars indicate SEM.
Cognitive Performance Following Acoustic Stimulation
For overnight memory consolidation, participants performed worse following a period of sleep relative to pre-sleep for both STIM (p = .02) and SHAM (p = .002). Consistent with recognition typically being reduced following sleep, participants recognized fewer correct matches for STIM (p = .04) and SHAM (p = .008) and had higher false alarm rates (STIM: p = .004, SHAM: p = .02) on the Paired Associates task. No difference in memory consolidation was observed following STIM compared with SHAM (p = .3; Table 4). Similarly, no group differences were observed between STIM and SHAM on any measure of executive function (p > .1).
Performance on Cognitive Tasks
| Task | Subtask | STIM | SHAM | P |
|---|---|---|---|---|
| Paired associates | Change score | −0.17 ± 0.07 | −0.21 ± 0.06 | .71 |
| Go No Go | d’ | 3.9 ± 0.18 | 3.75 ± 0.21 | .33 |
| N-back | 0-back d’ | 5.84 ± 0.12 | 5.84 ± 0.13 | .99 |
| 1-back d’ | 5.34 ± 0.25 | 5.36 ± 0.25 | .95 | |
| 2-back d’ | 3.75 ± 0.29 | 4.08 ± 0.31 | .23 | |
| Tower of London | Planning | 12.56 ± 1.22 | 13.55 ± 1.39 | .5 |
| Verbal fluency | Letter fluency | 37.13 ± 2.4 | 37.32 ± 2.35 | .37 |
| Category fluency | 38.92 ± 1.75 | 38.87 ± 2.15 | .8 | |
| Switch total | 13.46 ± 0.6 | 14.25 ± 0.62 | .26 |
| Task | Subtask | STIM | SHAM | P |
|---|---|---|---|---|
| Paired associates | Change score | −0.17 ± 0.07 | −0.21 ± 0.06 | .71 |
| Go No Go | d’ | 3.9 ± 0.18 | 3.75 ± 0.21 | .33 |
| N-back | 0-back d’ | 5.84 ± 0.12 | 5.84 ± 0.13 | .99 |
| 1-back d’ | 5.34 ± 0.25 | 5.36 ± 0.25 | .95 | |
| 2-back d’ | 3.75 ± 0.29 | 4.08 ± 0.31 | .23 | |
| Tower of London | Planning | 12.56 ± 1.22 | 13.55 ± 1.39 | .5 |
| Verbal fluency | Letter fluency | 37.13 ± 2.4 | 37.32 ± 2.35 | .37 |
| Category fluency | 38.92 ± 1.75 | 38.87 ± 2.15 | .8 | |
| Switch total | 13.46 ± 0.6 | 14.25 ± 0.62 | .26 |
RT, reaction time; SET, set execution time.
Performance on Cognitive Tasks
| Task | Subtask | STIM | SHAM | P |
|---|---|---|---|---|
| Paired associates | Change score | −0.17 ± 0.07 | −0.21 ± 0.06 | .71 |
| Go No Go | d’ | 3.9 ± 0.18 | 3.75 ± 0.21 | .33 |
| N-back | 0-back d’ | 5.84 ± 0.12 | 5.84 ± 0.13 | .99 |
| 1-back d’ | 5.34 ± 0.25 | 5.36 ± 0.25 | .95 | |
| 2-back d’ | 3.75 ± 0.29 | 4.08 ± 0.31 | .23 | |
| Tower of London | Planning | 12.56 ± 1.22 | 13.55 ± 1.39 | .5 |
| Verbal fluency | Letter fluency | 37.13 ± 2.4 | 37.32 ± 2.35 | .37 |
| Category fluency | 38.92 ± 1.75 | 38.87 ± 2.15 | .8 | |
| Switch total | 13.46 ± 0.6 | 14.25 ± 0.62 | .26 |
| Task | Subtask | STIM | SHAM | P |
|---|---|---|---|---|
| Paired associates | Change score | −0.17 ± 0.07 | −0.21 ± 0.06 | .71 |
| Go No Go | d’ | 3.9 ± 0.18 | 3.75 ± 0.21 | .33 |
| N-back | 0-back d’ | 5.84 ± 0.12 | 5.84 ± 0.13 | .99 |
| 1-back d’ | 5.34 ± 0.25 | 5.36 ± 0.25 | .95 | |
| 2-back d’ | 3.75 ± 0.29 | 4.08 ± 0.31 | .23 | |
| Tower of London | Planning | 12.56 ± 1.22 | 13.55 ± 1.39 | .5 |
| Verbal fluency | Letter fluency | 37.13 ± 2.4 | 37.32 ± 2.35 | .37 |
| Category fluency | 38.92 ± 1.75 | 38.87 ± 2.15 | .8 | |
| Switch total | 13.46 ± 0.6 | 14.25 ± 0.62 | .26 |
RT, reaction time; SET, set execution time.
Individual Differences in Slow-Wave and Cognitive Outcomes
Over 65% of participants exhibited SWE enhancement (>3% enhancement were considered responders), whereas 30% had a decrease (>3% reduction were considered nonresponders), and 4% were stable (0% ± 3%) (Figure 5A). Due to these inter-individual differences in enhancement, examining cognitive performance at the group level may mask any cognitive improvement from SWA enhancement. For this reason, we applied individual level examination of the cognitive responses of responders against nonresponders. Participants were ranked for percentage change in SWA. Nonresponders (n = 7) were defined as those who had no SWE improvement on the STIM night and compared against the remaining responders (n = 15). Responders had, on average, a 28.36% enhancement in SWE compared with an 11.8% decrease in nonresponders (STIM median: 19.51, SHAM median: −9.12, Mann–Whitney U = 0, n1 = 15, n2 = 7, p < .0001; Figure 5B). Again, significant differences for these groups were observed for absolute data (STIM: 16.99 ± 3.7, SHAM: −12.19 ± 4.9, t(20) = 5.9, p < .0001) (see Supplementary Figure 4A and B). We then compared these groups for cognitive outcomes.
Individual differences in slow-wave energy (SWE) between responders and nonresponders. (A) Percent change in relative SWE (STIM-SHAM/SHAM*100) for each individual and (B) all nonresponders (< −3% change; n = 7) compared with responders (>3% change; n = 15). Error bars indicate SEM.
Individual differences in slow-wave energy (SWE) between responders and nonresponders. (A) Percent change in relative SWE (STIM-SHAM/SHAM*100) for each individual and (B) all nonresponders (< −3% change; n = 7) compared with responders (>3% change; n = 15). Error bars indicate SEM.
For cognitive outcomes (expressed as percent change following STIM relative to SHAM), responders showed greater improvement on Verbal Fluency when compared with nonresponders, for phonetic fluency (t(18) = 2.37, p = .03, d = 1.3) (see Figure 6A). A comparable improvement in cognition was also observed for working memory, whereby responders showed significant improvement in 2-back d’, relative to nonresponders (t(20) = 2.35, p < .05, d = 1.2; Figure 6B). The same result was not observed for 1-back, likely due to ceiling effects for both conditions. When looking at individual performance, percentage improvement in performance was positively associated with the percentage of SWA enhancement for phonetic fluency (r = .68, p < .5, n = 17, with outliers [r = .14, p > .5, n = 21; Figure 6C], and working memory [r = .56, p < .01, n = 222, with outliers: r = .4, p = .056, n = 23]; Figure 6D). This was also observed for absolute data (see Supplementary Figure 5A–D). There were no significant differences between groups on measures of overnight memory consolidation (p > .6), inhibition (p > .2), or planning (p > .5).
Changes in slow-wave energy and cognition from STIM to SHAM for responders and nonresponders. Comparison of responders (n = 15) and non-responders (n = 7) responders on cognitive performance (top row) for (A) phonetic fluency and (B) 2-back d’. Significant positive associations were observed between magnitude of SWE change and cognitive improvement (bottom row) for (C) phonetic fluency and (D) 2-back d’. For (A) and (B), error bars indicate SEM. For (C) and (D), error is 95% confidence intervals of the regression line.
Changes in slow-wave energy and cognition from STIM to SHAM for responders and nonresponders. Comparison of responders (n = 15) and non-responders (n = 7) responders on cognitive performance (top row) for (A) phonetic fluency and (B) 2-back d’. Significant positive associations were observed between magnitude of SWE change and cognitive improvement (bottom row) for (C) phonetic fluency and (D) 2-back d’. For (A) and (B), error bars indicate SEM. For (C) and (D), error is 95% confidence intervals of the regression line.
Discussion
We describe the enhancement of SWA via a novel, automated acoustic stimulation device that can be readily deployed in at-home settings, in accordance with previous studies utilizing this specific stimulation method [33]. Moreover, we demonstrate that acoustic stimulation of SWA may lead to enhanced cognitive function, beyond sleep-dependent declarative memory, and during a stage of life marked by depleted SWA, and yet, requiring perhaps the largest cognitive capacity in the workplace. Our study therefore provides unique evidence for a wider use of acoustic stimulation within the general population to enhance SWA during sleep with an aim of improving cognitive function.
Our study in middle-aged men replicated existing literature showing SWA enhancement via acoustic stimulation in young [19] and older adults [7]. Although there was no change in the time spent in SWS (N3), SWA was enhanced due to an amplification of the slow oscillations. Taking into account previous studies of younger and older adults and notwithstanding the different methodological approaches, there does appear to be a dose–response association to the magnitude of the enhancement as a function of age. For instance, the percentage improvement in SWA was 15.6% (±5.6%) for younger adults [19], 14.3% (±5.82%) for middle-aged adults (our study), and 8% (±2%) for older adults [7]. Given the well-known depletion of SWS with age [49], and the need for the presence of SWS for the acoustic stimulation to be administered to the up phase of the slow oscillations, this would be expected. Future studies should confirm this with the use of the same experimental design and acoustic stimulation approach within a wide range of ages.
Previous studies have not shown an enhancement of SWA across the full-night spectra, but rather during STIM-ON periods versus STIM-OFF within the same sleep period (e.g. refs. [7, 19]). Contrary to this, we did report a general increase in SWA when comparing whole night spectra between STIM and SHAM. This could be driven by differing technologies [7], different analysis approaches (previous studies normalized to device-OFF periods, whereby we normalized to REM as an OFF period of device activation), or age group (although younger people appeared more responsive to SWA enhancement). Although we found relatively stable intraindividual differences in device activation (STIM vs. SHAM, albeit the latter with no sound), we did observe a wide range of interindividual responses to acoustic stimulation; this was also reported in an older age range [7], although less so in younger individuals [19].
We reported no significant improvement in cognitive outcomes at the group level (STIM vs. SHAM), which we attribute to the large individual differences in SWA enhancement (−22.04% to 106.38%). Responders however had a clear and significant improvement in multiple cognitive outcomes (phonetic fluency and working memory), compared with nonresponders. Moreover, the magnitude of SWA enhancement was positively associated with the magnitude of improvement in these executive functions: the more SWA was enhanced, the greater the improvement in next-day verbal fluency and working memory. Although we have previously demonstrated that <1 Hz SWA was linked to better performance on tasks of executive function [8], no causality could be attributed. The observation of experimental manipulation of SWA, either its depletion or enhancement, provides strong evidence of causality; thus, we provide further evidence of a causal role of SWA in these improved cognitive outcomes, which is in line with previous studies [5]. The benefit of SWS on next-day performance for cognition has been previously reported [50], which may be attributable to the reorganization of large-scale cortical networks [51] caused, in part, by the synaptic downscaling during SWS [52]. Contrary to previous studies assessing the cognitive impact of SWA enhancement, we found no improvement in sleep-dependent memory consolidation, either at the group level or at the individual level. We believe these null findings may be due to the task we used, as participants reported difficulty with the Paired Associates task (we did report lower memory retention rates than reported previously [5]). This may have been due to the use of nonsemantically related words, which is known to reduce the capacity for mnemonic strategies [53] and which differs to previous studies of acoustic stimulation which utilize tasks with semantically meaningful words [7, 19]. This enhanced difficulty in task completion is further supported by the lack of overnight improvement in both conditions (i.e. morning recall was always worse).
Our study is the first to specifically target middle-aged men. Previous studies had targeted either healthy, young adults (which is typical for proof of concept studies) [19] or older adults [7], given the association between SWA and memory decline in later life. Middle age, however, represents a clear target for SWA enhancement. Beyond the age of 35 years old, SWA rapidly depletes [10], at a rate of 2%–3% per decade. The fourth and fifth decade of life (30–50 years) however necessitates optimal cognitive capacity with respect to employment and productivity. Our data suggest that middle-aged adults, particularly those who rely on high executive function demands, such as flexible thinking, communication and updating of information, may maximize waking outcomes with this approach. In addition, the enhancement of SWA during sleep has been argued as a potential therapeutic and preventative strategy for reducing the risk of developing dementia caused by Alzheimer’s disease [54]. Although SWA does indeed improve memory function in healthy, cognitively intact older adults [7], the extent to which this approach would prevent or delay the development of cognitive pathologies is unknown. Given that the neurological indices of dementia can precede significant cognitive dysfunction (particularly memory) by over a decade [55], early intervention is essential and our data provide the first evidence of the capacity to enhance SWA in middle age. As society ages, and dementia risks continue in an upward trajectory (by 275% by 2056 in Australia [56] and 150% in the United States by 2060 [57]), acoustic stimulation may offer a viable solution to the Institute of Medicine’s recommendation for targets to improve cognitive outcomes in older adults [58]. As we have shown an independent validation of a commercially available device, SWA enhancement via acoustic stimulation represents an exciting modifiable sleep target for improved cognitive function in the wider community. Future studies should examine the utility of a long-term use of an easy-to-use, at-home acoustic stimulation device on cognitive function in older adults.
Our study has a number of limitations. First, and consistent with other proof of concept studies, we employed strict screening criteria to ensure our population was as homogenous as possible. The extent to which those with poorer sleep, sleep disorders, or daytime dysfunction would benefit from this approach remains unknown. Second, due to clear sex differences in SWA [59], we chose to focus only on middle-aged men as SWA depletion is heightened for men [60]. As such, our study findings may not be applicable to women. Third, normalizing SWA power spectra to REM may not be considered a typical approach; however, this was necessary for our study design. Although relative EEG power is typically derived relative to the full power spectra, this was not appropriate in our study (see Materials and Methods). Moreover, we were unable to examine within-night, epoch-by-epoch comparisons of STIM-ON to STIM-OFF, as per previous studies (e.g. refs. [7, 17, 19]) as ours was an independent evaluation of an automated device (i.e. we had no access to that data). We do note, however, that the device we used has previously been shown to enhance SWA in ON epochs relative to OFF epochs [33]. As we report no differences in REM power spectra, demonstrate stability in NREM SWA/REM SWA ratios across nights, and that the absolute data support our relative EEG data, we do not believe this approach accounted for the observed differences in relative SWA/SWE following STIM, as described here.
Beyond addressing these limitations, our study suggests several avenues for future research. First, future studies should consider the effects of acoustic SWS augmentation on cognitive domains beyond memory consolidation. Second, and as mentioned previously [7], as this technology moves toward repeated at-home use, future studies should examine the long-term effects of acoustic enhancement on cognition and physiology more broadly in ecologically valid settings. Finally, more research needs to be conducted to identify factors that determine interindividual responses to acoustic stimulation to optimize it as a viable option for SWS enhancement in the wider community. This is particularly important as it appears that acoustic stimulation may not be effective for all individuals. As mentioned above, repeated use of acoustic stimulation would also allow us to elucidate trait characteristics between responders and nonresponders, which is a necessary next step in determining the potential positive impact of acoustic stimulation for the wider community. With further research, particularly with devices that can be readily employed outside the confines of the laboratory, the future for SWA enhancement presents as an exciting opportunity for improving memory and executive function across the lifespan, and may prove to be a critical tool for preventing or slowing down cognitive decline in later life, particularly for dementia where decaying SWA may moderate the behavioral (memory decline) and biological (beta-amyloid, tau protein) antecedents of disease progression [5, 54, 61]. Until then, we present evidence for enhancing SWA using a novel, automated device, which has a beneficial impact on next-day cognitive outcomes in middle-aged men.
Supplementary material
Supplementary data are available at SLEEP online.
Figure S1. Relative full-night spectra for STIM and SHAM during baseline nights. No significant differences observed in relative power between conditions.
Figure S2. Comparison between STIM and SHAM of each sleep stage in minutes for each cycle. No significant differences were observed between conditions for each sleep stage.
Figure S3. Absolute data showing (A) differences in power in full-night spectra for STIM night relative to SHAM, in 0.5 Hz bins; (B) for each pre-determined EEG spectral bandwidth including delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), slow spindles (SS, 12–14 Hz), fast spindles (FS, 14–16 Hz), and beta (16–30 Hz); and (C) SWA in the delta range;
Figure S4. Percent change in absolute SWE (STIM-SHAM/SHAM*100) for (A) each individual and (B) all nonresponders (< −3% change; n = 9) compared to responders (n = 12).
Figure S5. Changes in absolute slow-wave energy and cognition from STIM to SHAM for responders and non-responders for (A) Phonetic fluency and (B) 2-back d’. Pearson’s correlations between SWE%Change and (C) Phonetic fluency and (D) 2-back d’. Error bars indicate SEM. Error is 95% confidence intervals of the regression line.
Funding
The study was supported by a Project Grant from the Cooperative Research Centre (CRC) for Alertness, Safety, and Productivity, Melbourne, Australia (grant number P3.1.07-17).
Author Contributions
All authors have made substantial contributions to the work presented and have approved the final version of the manuscript. C.A. and S.F. designed the study with input from SPAD. Both C.D. and S.F. were responsible for the collection of data; C.D., J.E.M., S.F., C.L.N., and C.A. analyzed the data; C.D. and C.A. interpreted the data, and C.D. wrote the manuscript with edits from C.A. All authors approved the final manuscript.
Conflict of Interest statement.
J.E.M., C.N., and S.P.A.D. report no competing financial interests. C.D. is a recipient of a PhD Scholarship, S.F. is a Project Leader, and C.A. is a Theme Leader in the Cooperative Research Centre for Alertness, Safety and Productivity. In the interest of full disclosure, C.A. has received a research award/prize from Sanofi-Aventis; contract research support from VicRoads, Rio Tinto Coal Australia, National Transport Commission, Tontine/Pacific Brands; and lecturing fees from Brown Medical School/Rhode Island Hospital, Ausmed, Healthmed, and TEVA Pharmaceuticals; and reimbursements for conference travel expenses from Philips Healthcare. In addition, she has served as a consultant through her institution to the Rail, Bus, and Tram Union, the Transport Accident Commission (TAC), and the National Transportation Committee (NTC). She has also served as an expert witness and/or consultant in relation to fatigue and drowsy driving.
Work was conducted at the Sleep and Circadian Medicine Laboratory, Turner Institute for Brain and Mental Health, Monash University, Melbourne, Victoria, Australia
Acknowledgments
We thank the study participants, technicians, staff, and students of the Monash University Sleep and Circadian Medicine Laboratory who aided data collection. We also thank Caroline Beatty for recruitment efforts and Christopher Andara for research assistance in the study. The study was approved by Monash University Human Research Ethics Committee.






