Slow Wave Sleep: What Deep Sleep Does, and Its Link to Growth Hormone

Slow wave sleep is the deepest stage of non-REM sleep, when the EEG fills with large, slow delta waves, the sleeper is hardest to wake and the pituitary releases its largest pulse of growth hormone. It also declines markedly with age, especially in men, which puts deep sleep at the centre of research on the growth hormone axis and the peptide signals that link the two.

This guide covers how sleep is staged and how it cycles, what drives slow wave sleep, how it changes with age, its tie to growth hormone, what peptide studies have shown and where the evidence stops.

The stages of sleep, as scored today

Sleep is staged from overnight recordings of brain waves (electroencephalography, EEG), eye movements and muscle tone. In 2007 the AASM replaced the older Rechtschaffen and Kales scheme of four non-rapid eye movement (non-REM) stages with three, N1, N2 and N3, plus REM. The old stages 3 and 4, which differed in how much high-voltage slow wave activity they showed, became N3.

N3’s signature is slow wave activity: waves of 0.5–2 Hz with a peak-to-peak amplitude above 75 µV, filling at least 20% of an epoch. The threshold is a convention, not a biological boundary. Slow waves shrink with age, and a 2024 review notes that older adults’ slow wave activity often misses the amplitude criterion, so those epochs are scored as N2.

Sleep cycles across the night

Non-REM and REM sleep alternate in cycles: in the Institute of Medicine’s summary, the first lasts about 70–100 minutes and later ones about 90–120. Slow wave sleep is concentrated in the first third of the night and REM sleep is longest in the last third, while slow wave activity is highest early and declines as the night goes on.

What drives slow wave sleep: the two-process model

The standard account of this pattern is the two-process model, first described by the Zurich sleep researcher Alexander Borbély in 1982. Process S is a homeostatic sleep pressure determined by preceding sleep and wakefulness. Process C is a circadian rhythm set by the body’s master clock.

Slow wave activity is the readout of Process S. It declines across the sleep period, and slow wave sleep is enhanced in recovery sleep after deprivation, by an amount that depends on how long the subject had been awake. So a study that changes prior wakefulness has changed slow wave sleep too. The circadian side, including night-time melatonin, is covered in our guide to what the pineal gland does.

How deep sleep changes with age

The Institute of Medicine’s summary puts the decline in slow wave sleep at about 2% per decade between ages 20 and 60, larger in men than in women.

The key hormonal data set pooled 149 healthy men aged 16 to 83. Slow wave sleep fell from 18.9% of the night at 16–25 to 3.4% at 36–50 and was replaced by lighter sleep. After midlife it fell no further, while time awake rose by about 28 minutes per decade. GH secretion fell steeply over the first transition and slowly afterwards, and slow wave sleep predicted GH secretion independently of age.

A 2019 meta-analysis using only the 2007 AASM rules found no significant age-related fall in slow wave sleep, and an earlier meta-analysis found no significant change after 60. Older adults’ smaller slow waves also often miss the 75 µV threshold and are scored as N2. And women generally have more slow wave sleep than men, yet most of the human studies in this article enrolled only men.

Sleep and growth hormone: the pulse after the first slow wave episode

In 1968, Takahashi, Kipnis and Daughaday followed eight young adults over 38 nights. In seven of them, a GH peak lasting 1.5 to 3.5 hours appeared with the onset of deep sleep; delaying sleep delayed the peak, and waking subjects for two to three hours before letting them sleep again produced another. In 1969, Sassin and colleagues showed that reversing the sleep–waking cycle by 12 hours reversed GH release with it: the pulse follows sleep rather than a clock of its own.

In 1991, Holl and colleagues sampled blood and scored sleep every 30 seconds for eight hours in six fasted young men. GH concentrations were higher in stages 3 and 4 than in other stages, and secretion peaked within minutes of their onset. The coupling is strong but not exclusive: the same data identified about 1.2 GH pulses per hour across the night. See also our guide to IGF-1 and the GH/IGF-1 axis.

The circuit behind it, in mice

In a 2025 mouse study from Yang Dan’s laboratory, somatostatin (SST) neurons in the arcuate nucleus suppressed GH release by inhibiting nearby GHRH neurons, and periventricular SST neurons acted at the median eminence. In non-REM sleep, GHRH activity rose moderately and SST activity fell; in REM sleep both surged, and GH release rose in both states, unlike the human data, where GH was higher in stages 3 and 4 than in REM. GH in turn excited locus coeruleus neurons and increased wakefulness, a negative feedback loop. See our guide to somatostatin, the brake on growth hormone.

GHRH, somatostatin and ghrelin: the peptide signals

Ferenc Obál and James Krueger’s group and Axel Steiger’s group in Munich proposed that one hypothalamic neuropeptide, growth hormone-releasing hormone (GHRH), links GH release and slow wave sleep; our growth hormone peptides research map shows where GHRH analogues and ghrelin mimetics sit.

Animal evidence

In 1988, GHRH given into the brain ventricles of rats and rabbits promoted non-REM and REM sleep and increased slow wave activity. In 2001, dwarf rats with defective GHRH signalling had less spontaneous non-REM sleep, and after four hours of sleep deprivation their slow wave activity rose only about half as much as in controls.

Small human studies

In 1992, Steiger’s group gave GHRH, somatostatin or placebo repeatedly in the early night to seven healthy men. Slow wave sleep made up 20.2% of the night after GHRH against 14.0% after placebo, and nocturnal cortisol was blunted; short-acting somatostatin changed neither. Seeing the GHRH effects as the mirror image of corticotropin-releasing hormone’s, the authors proposed that GHRH and CRH interact to regulate sleep (see our guide to ACTH and the HPA axis). In 2004 the same group gave the long-acting somatostatin analogue octreotide to seven young men: stage 4 sleep fell across the night and waking increased in the second half, suggesting that GHRH and somatostatin regulate sleep reciprocally.

Ghrelin and GHRP-6

Ghrelin was tested the same way: in seven men, nights with repeated early-night ghrelin showed more slow wave sleep and less REM sleep in the second third of the night (see our ghrelin explainer). The synthetic hexapeptide GHRP-6 did not follow suit. In a 1995 study of healthy men it lengthened stage 2 sleep, left slow wave sleep unchanged and raised ACTH and cortisol, the opposite of GHRH’s effect on cortisol. The authors took the divergence as a sign that GHRPs and GHRH act through different receptors, a distinction our secretagogue explainer covers.

DSIP is this field’s cautionary tale. In 1977 a Basel group infused material dialyzed from the cerebral venous blood of thalamically stimulated rabbits into other rabbits, recorded more cortical delta activity and traced it to a nine-residue peptide. A 2006 review called it “a still unresolved riddle”: no gene, precursor or receptor had been isolated, and in the reviewers’ early studies DSIP itself did not promote slow wave sleep. Our DSIP research guide has the history.

What deep sleep is linked to

Memory and the slow oscillation

In 2006, Marshall, Born and colleagues applied transcranial stimulation oscillating at 0.75 Hz during early-night non-REM sleep in healthy volunteers. It increased slow wave sleep and slow oscillations, the sub-1 Hz component of slow wave activity, and improved retention of hippocampus-dependent declarative memories; 5 Hz stimulation did not. The authors concluded that slow oscillations contribute causally to sleep-associated memory consolidation.

Brain clearance: a contested mouse story

A 2013 mouse study from Maiken Nedergaard’s group found that sleep and anesthesia expanded the brain’s interstitial space by about 60% and sped the clearance of beta-amyloid. A 2024 study in male mice, from a group including Nicholas Franks, found clearance of fluorescent molecules markedly lower during sleep and anesthesia than during wakefulness. In people, a 2019 imaging study showed non-REM slow waves followed by hemodynamic oscillations coupled to waves of cerebrospinal fluid flow, which links slow waves to fluid movement but does not measure clearance. The question is open.

Metabolism: insulin sensitivity

In 2008, Tasali, Van Cauter and colleagues suppressed slow wave sleep in young, healthy adults without changing their total sleep time. Insulin sensitivity fell without adequate compensation, glucose tolerance dropped, and the loss tracked the reduction in slow wave sleep: a short experiment, not proof of long-term risk.

Muscle: what one night without sleep did

In a 2021 randomized crossover study of 13 healthy young adults, seven men and six women, one night without sleep lowered post-meal muscle protein synthesis by 18%, raised plasma cortisol by 21% and lowered testosterone by 24%, with no change in markers of breakdown. It measured total sleep loss, not slow wave sleep, and tested no compound.

What’s established and what isn’t

Sourcing growth-axis research peptides in Canada

For Canadian labs studying GHRH-receptor or secretagogue-receptor signalling in vitro, the documentation questions are those for any research peptide; our overview of research peptides in Canada sets them out.

• Check identity and purity separately. Mass spectrometry confirms the expected mass; HPLC shows how much of the sample is that molecule. Our peptide COA guide explains both.
• Match the batch number on the certificate to the vial.
• Know the molecule. Sermorelin is GHRH(1-29) amide, GHRP-6 a synthetic hexapeptide and DSIP the 1977 rabbit nonapeptide. Our sermorelin research guide compares GHRH analogues.
• Store lyophilized material cold, dry and dark; see the peptide storage guide.

SermorelinGHRH(1-29) amide, the reference analogue in GHRH-receptor researchView product GHRP-6Synthetic hexapeptide used in secretagogue-receptor researchView product DSIPNonapeptide first isolated from rabbit blood in 1977View product

Frequently asked questions