Sleep and Longevity

Sleep is not optional downtime. It is active biological maintenance. While you lie there, your body is repairing, clearing, and consolidating.

What happens while you sleep.

Brain clearance. The glymphatic system is a drainage network first mapped by Iliff and Nedergaard. It moves cerebrospinal fluid (CSF) through brain tissue and washes out metabolic waste, including beta-amyloid. The original 2013 mouse paper (Xie, Science) found the space between brain cells expands by roughly 60 percent during sleep, doubling the clearance rate of injected tracers. Popular media often turned this into "10x more active," which overstates the primary finding. Human MRI work (a 2019 Science paper on coupled EEG, BOLD, and CSF oscillations; the Eide and Ringstad DCE-MRI series) has shown CSF dynamics during sleep but has not directly quantified Aβ/tau clearance. A 2024 Nature Neuroscience paper from the Franks lab used a different injection method and reported reduced clearance during sleep. The field is in active dispute. Honest summary: CSF dynamics change during sleep, but the size of the effect depends on what you measure.

Memory consolidation. Information moves from short-term to long-term storage. Slow-wave sleep (N3, the deepest non-REM stage with high-amplitude slow EEG waves) handles declarative memory through hippocampal sharp-wave ripples coupled to thalamocortical spindles and cortical slow oscillations. REM (rapid eye movement sleep, when most vivid dreams happen) handles procedural and emotional memory. Researchers like Walker, Stickgold, Diekelmann and Born have spent two decades mapping this active systems consolidation model.

Hormone regulation. Growth hormone (GH) peaks during N3. Roughly two-thirds of daily GH output happens during nocturnal sleep, locked to the first deep-sleep episode. Sleep also regulates appetite hormones (leptin and ghrelin), cortisol (which rises before you wake, the cortisol awakening response), and prolactin.

Immune function. Your immune system makes and releases cytokines and antibodies during sleep. Chronic shortfall measurably weakens vaccine responses and viral resistance.

DNA repair and cellular maintenance. Repair processes peak during sleep across nearly every cell type studied.

What this means for lifespan. A meta-analysis of 1.3 million people (Cappuccio, Sleep, 2010) found that short sleep (≤6 hours) and long sleep (≥9 hours) were both linked to higher mortality risk: 12 percent and 30 percent respectively. The curve is U-shaped. Long sleep may partly reflect underlying illness rather than causing harm directly. Several widely-cited causal claims in popular sleep writing have been challenged (Guzey's 2019 critique of Why We Sleep). Treat the Cappuccio curve as the baseline evidence.

A normal night is not uniform. Your brain cycles through four distinct EEG stages (N1, N2, N3, and REM) in roughly 90-minute bouts.

The stages. N1 (~2 to 5 percent of total sleep time in young adults) is the drift from wake into sleep, with low-amplitude theta EEG and slow rolling eye movements. N2 (~45 to 55 percent) is the workhorse stage, defined by sleep spindles (11 to 16 Hz bursts from the thalamic reticular nucleus, the brain's sensory gate-keeper) and K-complexes. N3 is slow-wave sleep, the deep stuff. It is scored when ≥20 percent of an epoch holds slow waves at ≤2 Hz and ≥75 µV. In healthy young adults, N3 is roughly 13 to 23 percent of total sleep. REM (~20 to 25 percent) is paradoxical sleep, with desynchronised EEG that resembles wake, rapid eye movements, hippocampal theta, and skeletal muscle atonia (your body is briefly paralysed so you don't act out your dreams).

The pattern. Cycles last roughly 90 minutes (range 70 to 120) and repeat four to six times per night. The structure is lopsided. SWS dominates the first two cycles and is often absent from the last. REM lengthens across the night, and the longest REM bout usually lands just before morning waking. Cut sleep short on the back end and you lose mostly REM. Cut it short on the front end and you lose mostly SWS. The popular "wake at the end of a 90-minute cycle" advice oversimplifies. Cycle length varies 30+ minutes between and within people.

The two-process model. Borbély's two-process model (Human Neurobiology, 1982) describes sleep regulation as the interaction of Process S (homeostatic sleep pressure that builds up while you're awake) and Process C (circadian alerting signal from the suprachiasmatic nucleus, the master clock in your brain). Sleep happens when S crosses a threshold set by C. The molecular driver of Process S is best characterised as adenosine (a tiredness-signal molecule that piles up while you're awake). A 1997 Science study (Porkka-Heiskanen) showed extracellular adenosine rises in the basal forebrain during long wakefulness and falls during recovery sleep. Caffeine blocks adenosine A1 and A2A receptors. It doesn't add energy. It blocks the tiredness signal.

Sleep architecture changes with age. A 2004 Sleep meta-analysis (Ohayon) pooled 65 studies covering ages 5 to 102. In adults, total sleep time falls ~10 minutes per decade, sleep latency rises, N1 and N2 percentages rise, SWS percentage falls, REM percentage falls, and wake after sleep onset rises. A 2000 JAMA paper (Van Cauter) documented the SWS collapse: from ~19 percent of sleep at ages 16 to 25 down to ~3.4 percent at ages 36 to 50 in men. That's roughly 38 minutes per decade lost through midlife, then it stabilises. Nocturnal GH secretion fell ~75 percent in parallel.

Here is the practical part: the "eight hours of consolidated SWS-rich sleep" picture describes young adults. By 60, fragmentation is real. Protecting whatever N3 you have left becomes the longevity-relevant target. See our deep sleep guide for the SWS-specific levers.

Poor sleep speeds up biological aging through several paths.

Telomere shortening. Chronic sleep loss is linked in observational studies to shorter telomeres (the protective caps on your chromosomes that get shorter every time a cell divides). Effect sizes vary. The direction is consistent.

Epigenetic shifts. Sleep disruption changes DNA methylation patterns, the same markers used in biological-age tests. A 2018 Science Advances study (Cedernaes) showed that one night of sleep deprivation alters methylation at clock genes and lowers OXPHOS transcripts in skeletal muscle.

Inflammation. Sleep loss raises CRP, IL-6, and TNF-α (three blood markers of inflammation). This feeds "inflammaging," the slow-burn chronic inflammation that drives age-related disease.

Insulin resistance. One week of 5-hour sleep restriction in healthy young adults dropped whole-body insulin sensitivity by 11 to 20 percent (Buxton, Diabetes, 2010). Earlier, more severe restriction (4 hours for 6 nights, Spiegel, Lancet, 1999) cut glucose tolerance by 30 to 40 percent, which is pre-diabetes territory.

Cognitive aging. The Whitehall II cohort (Sabia, Nature Communications, 2021) followed ~8,000 British civil servants for 25 years. People who habitually slept ≤6 hours at age 50 had a 22 percent higher risk of later dementia. Your brain needs sleep to clear Alzheimer's-associated proteins.

The link runs both ways. Aging changes sleep on its own. Older adults get less SWS and more fragmentation. So poor sleep speeds aging, and aging worsens sleep. A loop that feeds itself.

The upside. Sleep quality is not fixed. Better sleep habits, treatment of sleep disorders, and the lifestyle changes in this guide can slow these aging effects.

The short answer: 7 to 9 hours for adults aged 18 to 64, and 7 to 8 hours from age 65 onward. That is the consensus across the major sleep medicine societies, and it is the range to anchor on.

The sweet spot. Recent large cohorts (Li et al. 2022 Nature Aging using UK Biobank) point to around 7 hours as the midpoint tied to lower all-cause mortality and better thinking in middle-aged and older adults. 7 to 8 hours is safe for most. Both shorter and longer come with more health risk.

Quality over quantity. Sleep isn't just time in bed.

• Sleep efficiency: time asleep ÷ time in bed. Aim for ≥85 percent.
• Sleep stages: you need enough N3 for physical recovery and enough REM for memory and thinking.
• Sleep continuity: unbroken sleep restores you more than the same hours chopped into pieces.

Signs your sleep is good enough:

• You wake without an alarm feeling rested
• You hold energy through the day without leaning on caffeine
• You fall asleep in 15 to 20 minutes
• You don't wake up often at night
• You don't feel wiped out during the day

Signs it isn't:

• You need an alarm to get up
• You rely on caffeine to function
• You crash in the afternoon
• You fall asleep the instant your head hits the pillow (often a sign of sleep debt)
• You feel drowsy driving or in meetings

Your 2-week sleep diary. Before changing anything, measure. For 14 days, track: bedtime, wake time, caffeine cutoff, alcohol units, perceived sleep quality 1 to 5. Any notebook works. A wearable with a sleep log works too.

Two numbers to compute at the end:

• Sleep efficiency: time asleep ÷ time in bed. Target ≥85 percent.
• Sleep latency: time to fall asleep. Healthy is 15 to 30 min. Under 5 min = likely sleep debt. Over 30 min = onset insomnia.

Your sleep timing is set by light and temperature, not by willpower.

The master clock. The suprachiasmatic nucleus (SCN, a small cluster of neurons in the hypothalamus that runs your 24-hour rhythm) reads time from your retina via melanopsin-expressing intrinsically photosensitive retinal ganglion cells (ipRGCs, special non-vision cells in the eye tuned to blue light), with peak sensitivity at ~480 nm (short-wavelength blue). A 2002 Science study (Berson) pinned down the receptor cells. A 2003 Journal of Physiology paper (Khalsa) mapped the human phase-response curve: a 6.7 h bright light pulse can advance the clock up to 2.0 h or delay it up to 3.6 h. The system delays more easily than it advances.

Morning light is the most important variable. Outdoor light on a clear day is 10,000 to 100,000 lux at the eye. A bright office is 300 to 500 lux. A typical living room in the evening is 50 to 200 lux. The circadian system reacts to brightness in a non-linear way, and a typical indoor space sits in the "barely registers" zone for clock-setting. A 2013 Current Biology study (Wright) sent volunteers camping for a week. Daytime light exposure rose fourfold, the biological night realigned to the natural light-dark cycle, and the group's chronotype distribution shifted closer to solar time. The expert consensus (Brown, PLOS Biology, 2022) recommends a minimum melanopic EDI of 250 lux at the eye during the day.

Practical: get outside within 30 to 60 minutes of waking, for 10 to 30 minutes, ideally without sunglasses. Window glass cuts the dose roughly in half versus stepping outside. A cloudy day still beats indoor lighting by 1 to 2 orders of magnitude.

Evening light is the other side. A 2000 Journal of Physiology study (Zeitzer) showed nighttime light suppresses melatonin (the hormone that signals "biological night" to your body) along a curve. About 50 percent suppression hits at ~100 lux of room light, with detectable suppression below 30 lux. A 2015 PNAS study (Chang and Czeisler) showed 4 hours of evening eReader use delayed melatonin onset by ~1.5 hours and hurt next-morning alertness for hours after the light went off. The Brown 2022 consensus: melanopic EDI ≤10 lux at the eye during the 3 hours before bed, ≤1 lux during sleep. Most modern living rooms blow past 10 melanopic EDI from overhead LEDs alone. Blue-blocker glasses help by ~50 percent at typical screen exposure, but dimming the room overall works better.

Temperature gates sleep onset. Core body temperature falls roughly 0.5 to 1.0 °C across the night, reaching its low point around 04:00 to 05:00. The fall is driven by active heat loss from the body's surface, via distal vasodilation of hands and feet (your peripheral blood vessels open up and dump heat). A 2000 American Journal of Physiology study (Kräuchi) showed the distal-proximal skin temperature gradient (DPG, the temperature difference between your hands/feet and your trunk) is the single best predictor of sleep-onset latency. Better than core body temperature, better than melatonin, better than how sleepy you say you are. Warm extremities are the signal that sleep is coming.

Bedroom temperature. A 2012 Journal of Physiological Anthropology review (Okamoto-Mizuno and Mizuno) pooled the literature: 16 to 19 °C (60 to 67 °F) under typical bedding is the consensus range for adults. Heat is worse than cold. Above 26 °C with humidity, both SWS and REM decline. Cool the room, warm the extremities (socks help if you run cold).

Chronotype is real and largely genetic. Twin studies put heritability at 40 to 50 percent. Roenneberg's 2012 Current Biology work showed social jetlag (the gap between sleep timing on free vs. work days, basically how badly your weekday alarm clock disagrees with your biology) predicts metabolic outcomes. Every hour of social jetlag corresponded to ~33 percent higher odds of being overweight in the overweight subsample. You can shift chronotype 1 to 2 hours with disciplined light/dark scheduling, but you can't override it.

The substances most people put into their bodies in the evening are also the ones most likely to wreck their sleep.

Caffeine: the most underestimated. Half-life ranges 2 to 10 hours across CYP1A2 phenotypes (the liver enzyme that clears caffeine; some people are genetically fast metabolisers, others slow). A 2013 Journal of Clinical Sleep Medicine study (Drake) found 400 mg of caffeine taken 6 hours before bed cut total sleep time by more than an hour versus placebo. The "cut at 2 PM" rule only works if you go to bed at 10 PM and clear caffeine at the population average. Slow metabolisers do better with a noon or morning-only cutoff. Decaf is not zero: a 2006 Journal of Analytical Toxicology analysis (McCusker) measured 3.0 to 15.8 mg caffeine per shot of decaf espresso.

Alcohol: sedation first, fragmentation later. A 2013 Alcoholism: Clinical and Experimental Research review (Ebrahim) summarised it: alcohol shortens sleep onset and consolidates the first half of the night with deeper NREM. In the second half, sleep gets fragmented and arousals multiply as blood alcohol falls. REM is suppressed in the first half and rebounds in the second, often with vivid dreams. Wearable data routinely show high resting heart rate and depressed HRV (heart-rate variability, a measure of autonomic recovery) for 24 to 48 hours after even one or two drinks. Obstructive sleep apnea gets worse (reduced upper-airway muscle tone). Rule: finish ≥3 hours before bed, ≤1 drink on sleep nights, none if you have OSA.

Nicotine: a stimulant pretending to be a relaxant. Cholinergic stimulant pharmacology (it activates the same receptors as acetylcholine, your brain's alertness signal). It increases sleep latency, fragments sleep, reduces total sleep time and N3, and suppresses REM (Jaehne, Sleep Medicine Reviews, 2009). It worsens restless legs and increases periodic limb movements. Vaping evidence is thinner but lines up with the smoking literature. Rule: last nicotine ≥3 hours before bed. Quitting is the only path back to normal sleep architecture, and expect 2 to 4 weeks of disrupted sleep during the quit.

Late food and large meals. A 2019 Journal of Clinical Sleep Medicine study (Lopes) of 296 OSA patients found late dinner timing significantly raised AHI (apnea-hypopnea index, breathing pauses per hour; β = 1.28 events/h), increased sleep latency and wake-after-sleep-onset, and reduced REM. Why? The heat your body makes after eating blunts the nocturnal core-temperature fall (which gates sleep onset). Lying down with a full stomach worsens reflux and apnea. And overnight glucose spikes raise sympathetic tone. Rule: last meal ≥3 hours before bed; aim for a 10 to 12 hour eating window.

Cannabis: THC's trade and the withdrawal tax. A 2017 Current Psychiatry Reports review (Babson) laid it out: acute THC shortens sleep latency, may briefly increase SWS at low doses, and suppresses REM. Tolerance builds to the sleep-promoting effects. Withdrawal causes insomnia and reliably brings vivid dreams (REM rebound, when REM bounces back hard after suppression). A recent placebo-controlled high-density EEG trial of oral THC/CBD (Suraev, Sleep, 2023) reported a significant 8 percent reduction in REM percentage and 66-minute increase in REM latency. CBD alone, at anxiolytic doses, does not acutely disrupt sleep architecture (Linares, Frontiers in Pharmacology, 2018). Rule: nightly THC for sleep is a tolerance trap. If used, low dose and intermittent. Plan for 1 to 2 weeks of disrupted sleep on withdrawal.

Sleeping pills: Z-drugs out, DORAs in. The 2017 AASM clinical practice guideline (Sateia, Journal of Clinical Sleep Medicine) gave weak recommendations across all hypnotic classes and stressed CBT-I as first-line. Z-drugs (zolpidem, zaleplon, eszopiclone) and benzodiazepines suppress SWS, build tolerance, cause falls and complex sleep behaviours, and carry observational mortality signals (Kripke, BMJ Open, 2012: HR 3.6 to 5.3 across dose tertiles). In April 2019 the FDA added a boxed warning to zolpidem, zaleplon, and eszopiclone for complex sleep behaviours (sleep-walking, sleep-driving, and other activities while not fully awake), after reports of serious injuries and death. DORAs are dual orexin receptor antagonists (they block orexin, the wake-promoting brain signal). The class includes suvorexant, lemborexant, and daridorexant. They block orexin-driven wakefulness and preserve REM and SWS architecture. The phase 3 daridorexant trials (Mignot, Lancet Neurology, 2022; 1,854 patients across two trials at 1 and 3 months) showed efficacy on objective WASO and latency, with daytime-functioning improvement and a favourable safety profile. Rule: CBT-I first. If you need pharmacotherapy, prefer a DORA over a Z-drug, especially for chronic use or in older adults.

Common medications that distort sleep. Beta-blockers (especially lipophilic ones like propranolol, metoprolol, which cross into the brain) suppress nocturnal melatonin and cause vivid dreams and insomnia. A small RCT (Scheer, Sleep, 2012) showed 2.5 mg melatonin restored sleep quality in beta-blocker patients. SSRIs roughly double REM latency and suppress total REM (Wilson and Argyropoulos, Drugs, 2005). Corticosteroids cause insomnia, especially with evening dosing. Stimulant ADHD medications delay onset if dosed after noon. Rule: if insomnia starts within weeks of a new medication, suspect the medication. Move dosing to morning where clinically appropriate.

Four behaviours move the same lever (core body temperature) and drive most of the non-pharmacological sleep gains people chase. The evidence ranks them clearly.

Exercise modestly but reliably improves sleep. A 2019 Sports Medicine meta-analysis (Stutz) of 23 evening-exercise studies found evening exercise raised SWS by +1.3 percentage points (p=0.041), increased REM latency by 7.7 min, and cut N1 by 0.9 pp. The old "no evening exercise" rule is dead. Only vigorous exercise ending within 60 minutes of bed hurts sleep. Anything finishing ≥1 hour before lights-out is net neutral to positive, including for SWS.

Modality matters. Aerobic endurance produces the most consistent SWS increase. The mechanism is heat dumping, which triggers a core temperature drop. Resistance training improves subjective sleep quality (PSQI, a validated 7-item questionnaire) more than measured SWS (Kovacevic, Sleep Medicine Reviews, 2018). HIIT timing matters most: sub-maximal HIIT ≥90 minutes before bed is fine. Maximal-intensity work too close to bedtime delays onset.

Timing. Morning exercise pulls the circadian clock forward by ~0.6 hours per session (Youngstedt, Journal of Physiology, 2019, the exercise phase-response curve). That's useful for delayed sleep phase, eastward jet lag, or anyone who wants an earlier sleep window. Evening exercise (≥60 minutes before bed) doesn't disturb sleep and modestly boosts SWS.

Hot bathing is the most powerful thermal lever. A 2019 Sleep Medicine Reviews meta-analysis (Haghayegh) pooled 13 trials. Water at 40 to 42.5 °C (104 to 108.5 °F) for ≥10 minutes, 1 to 2 hours before bed, cut sleep onset latency by ~36 percent (Cohen's d ≈ 1.01) and improved sleep efficiency. The counter-intuitive mechanism: warming the periphery triggers reflexive vasodilation. When you step into a cool room, that dilated peripheral circulation dumps heat fast, and core temperature falls sharply over the next 60 to 90 minutes. The core-temperature decline is the sleep signal.

Hot-bath protocol: 40 to 43 °C water, 10 to 15 minutes, finish 60 to 120 minutes before lights-out, into a cool dim room.

Sauna. Strong cardiovascular evidence (the Finnish KIHD cohort, Laukkanen, BMC Medicine, 2018: 4 to 7 sessions/week linked with ~50 percent lower cardiovascular mortality vs. 1/week) but thin direct sleep RCT evidence. Mostly subjective reports and a single 1976 PSG study (n=5) showing +70 percent SWS in the first 2 hours. Mechanism overlaps with the hot bath. A reasonable evening protocol: 15 to 25 min at 80 to 90 °C, 1 to 2 hours pre-bed, cool (not cold) shower after.

Cold exposure. Cold-water immersion at ~14 °C produces a 200 to 300 percent norepinephrine spike (Šrámek, European Journal of Applied Physiology, 2000). Useful at 7 AM, a disaster at 9 PM. Morning or midday cold is fine. Pre-bed cold (within 1 hour) is alerting and delays onset. The Finnish sequence (sauna, then cool shower, then bed) is reasonable. Keep it to a brief 30 to 60 s cold exposure rather than deep immersion.

Cooling mattresses and the bedroom. A 2008 Brain study (Raymann and Van Someren) showed gentle warming of distal skin (~0.4 °C) doubled SWS from 8 to 14 percent in older adults. The independent Herberger 2024 Scientific Reports crossover (n=72) showed a high-heat-capacity mattress added +7.5 minutes of N3 and lowered heart rate ~2.4 bpm. Eight Sleep and chiliPad manufacturer-funded studies report larger effects but lack independent replication. A cool bedroom (17 to 20 °C) gets you most of the gain at no cost.

If you treat chronic insomnia with sleeping pills, you are treating it wrong. Every major sleep medicine society (AASM, the European Sleep Research Society, NICE, the American College of Physicians) now recommends Cognitive Behavioral Therapy for Insomnia (CBT-I) as first-line. Hypnotics are second-line, short-term, and lose effect fast.

The evidence. A 2015 Annals of Internal Medicine meta-analysis (Trauer) pooled 20 RCTs (1,162 patients): CBT-I cut sleep onset latency by 19 minutes, wake after sleep onset by 26 minutes, and raised sleep efficiency by 9.9 percentage points. The gains held at follow-up. A 2012 BMC Family Practice review (Mitchell) found CBT-I matches benzodiazepines acutely and beats them long-term. A 2017 Sleep trial (Beaulieu-Bonneau and Morin) reported, at 24-month follow-up of 160 patients, remission rates of 44 to 63 percent across CBT-I conditions. Patients who tapered off zolpidem while continuing CBT-I did better than those who kept the drug.

The five components. CBT-I is a structured, time-limited protocol, typically 4 to 8 sessions, built from:

1. Stimulus control (Bootzin 1972). Re-pair the bed with sleep, breaking its link to arousal and frustration.
2. Sleep restriction therapy (Spielman 1987). Deliberately reduce time in bed to consolidate fragmented sleep.
3. Cognitive restructuring. Test and revise catastrophic beliefs about sleep.
4. Relaxation training. Progressive muscle relaxation, paced breathing.
5. Sleep hygiene education. Caffeine, alcohol, light, exercise, temperature.

Here's the catch: sleep hygiene is the weakest component. The 2021 AASM guideline (Edinger) gives a conditional recommendation against sleep hygiene as standalone therapy. The active mechanism is sleep restriction plus stimulus control.

Sleep restriction is the engine. Counter-intuitive but most powerful. The protocol:

1. Keep a sleep diary for 7 to 14 days. Calculate average total sleep time (TST).
2. Set time in bed (TIB) = TST + 30 minutes, never less than 5 hours. Pick a fixed wake time first, then back-calculate bedtime.
3. Hold the window for 7 days. You will be sleepy. That's the point.
4. Titrate weekly: if sleep efficiency ≥85 percent, extend bedtime by 15 to 30 minutes earlier. If 80 to 84 percent, hold. If <80 percent, contract by 15 minutes.
5. Keep going until you reach a sustainable window (usually 6.5 to 7.5 hours) with high efficiency.

A 2011 Archives of Internal Medicine trial (Buysse) tested a 4-session brief behavioural protocol (BBTI) in older adults with chronic insomnia: 67 percent response, 55 percent remission, NNT 2.4 (number needed to treat, the patients you need to treat for one extra success).

Stimulus control rules.

• Go to bed only when sleepy (heavy eyelids, not just tired).
• Use the bed only for sleep and sex.
• If you don't fall asleep within ~20 minutes, get up. Do something quiet and dim, then return when sleepy.
• Same wake time every day, including weekends.
• No daytime naps during the protocol.

Digital CBT-I works. A 2020 Lancet Digital Health trial (Vedaa) randomised 1,721 Norwegian adults: digital CBT-I produced a Cohen's d of −1.21 on the Insomnia Severity Index vs. patient education, with 58 percent clinically meaningful response. Germany's DiGA program reimburses the somnio app on statutory insurance (GKV) prescription. NICE TA922 (2023) mandates CBT-I before daridorexant in the UK.

When to escalate to a specialist: ≥6 to 8 weeks of structured CBT-I at home without response; ISI score still >15 at week 8; significant daytime impairment; coexisting OSA, RLS, RBD, current major depression with suicidal ideation, shift work, recent Z-drug use needing a taper.

Not every sleep problem is fixable with hygiene and CBT-I. Some need a clinician.

Obstructive sleep apnea (OSA, when your airway repeatedly collapses during sleep). Prevalence in adults with AHI ≥15 (Peppard, American Journal of Epidemiology, 2013):

• Men 30 to 49: ~10 percent
• Men 50 to 70: ~17 percent
• Women 30 to 49: ~3 percent
• Women 50 to 70: ~9 percent

Risk factors: obesity, neck circumference, male sex, age, retrognathia (a recessed lower jaw).

Screen with STOP-BANG: snoring loud, tired daytime, observed apneas, hypertension, BMI >35, age >50, neck >40 cm, male sex. ≥3 = intermediate risk; ≥5 = high risk. Diagnosis is polysomnography (in-lab, the full overnight sleep study with EEG, breathing, and movement sensors) or home sleep apnea testing (HSAT, a portable version) for uncomplicated suspected moderate-severe cases.

Treatment ladder: CPAP (continuous positive airway pressure, the mask that splints your airway open with pressurised air) is the gold standard. Mandibular advancement devices for mild-moderate or CPAP-intolerant. Positional therapy. Weight loss. Hypoglossal nerve stimulation (Inspire, an implantable device that pulses the tongue-nerve to keep the airway open; the STAR trial, NEJM, 2014: AHI fell from 29 to 9 at 12 months). The newest addition is SURMOUNT-OSA (Malhotra, NEJM, 2024): tirzepatide reduced AHI by 20 to 24 events/hour beyond placebo in obesity-related OSA over 52 weeks. First disease-modifying drug in this space.

The SAVE (McEvoy, NEJM, 2016) and ISAACC (Sánchez-de-la-Torre, Lancet Respiratory Medicine, 2020) trials were null on cardiovascular endpoints, but mean CPAP adherence in both was only 2.8 to 3.3 hours per night, and both enrolled non-sleepy patients (low ESS), the subset least likely to benefit on patient-reported outcomes. CPAP remains first-line for symptomatic OSA.

Insomnia disorder. Covered in the CBT-I section. ≥3 nights/week × ≥3 months × daytime impairment × adequate sleep opportunity.

Restless legs syndrome (RLS / Willis-Ekbom disease). Urge to move the legs, worse at rest, relieved by movement, evening/night predominant. Always check ferritin and transferrin saturation (the two main iron storage and transport markers). The 2024 AASM guideline (Winkelman) gives a strong recommendation for IV ferric carboxymaltose in patients with appropriate iron status (in many protocols this means ferritin <75 ng/mL with TSAT <45 percent). Alpha-2-delta ligands (gabapentin enacarbil, gabapentin, pregabalin) are preferred first-line. Dopamine agonists (pramipexole, ropinirole), once first-line, are now recommended against for long-term use due to augmentation (paradoxical worsening of symptoms over time at ~7 percent per year cumulative).

REM sleep behavior disorder (RBD). Loss of REM atonia, dream enactment (kicking, punching, vocalising, jumping out of bed). This is the most clinically urgent red flag in this guide. A 2019 Brain study (Postuma; international cohort, 1,280 patients across 24 centres) documented a 6.3 percent per year phenoconversion rate to Parkinson's disease, dementia with Lewy bodies, or multiple system atrophy, reaching 73.5 percent within 12 years. RBD is the strongest known prodromal marker for alpha-synucleinopathy (the protein-misfolding family of disorders that includes Parkinson's). Refer to neurology, not just for neurologist-prescribed symptomatic management (clonazepam or melatonin 3 to 12 mg at night) but for prodromal monitoring and access to neuroprotective trial enrolment.

Circadian rhythm sleep-wake disorders. Delayed sleep-wake phase disorder (common in adolescents and young adults, 5 to 16 percent prevalence). DSWPD is a diagnosed condition. See a sleep clinician (Schlafmediziner) before starting. Standard treatment protocols use morning bright light plus low-dose melatonin (0.3 to 0.5 mg) taken 5 to 7 hours before habitual sleep onset, targeting the dim-light melatonin onset (DLMO, the moment your body starts releasing melatonin in the evening). Advanced sleep-wake phase, shift work disorder, and totally-blind N24SWD have their own protocols.

Narcolepsy. Type 1: orexin/hypocretin deficiency + cataplexy (sudden muscle weakness triggered by emotion). Type 2: no cataplexy, normal hypocretin. Diagnosis is polysomnography + multiple sleep latency test (mean sleep latency ≤8 minutes + ≥2 sleep-onset REM periods). HLA-DQB1*06:02 is present in >95 percent of type 1.

Periodic limb movements of sleep (PLMS): a PSG finding, treat only if PLMD (clinical disturbance). Same iron-first pathway as RLS.

Red flags: when to refer.

• Loud snoring + witnessed apneas + daytime sleepiness → OSA workup
• Dream enactment / acting out dreams → RBD → neurology referral
• Insomnia ≥3 months → CBT-I, not chronic Z-drug
• Excessive daytime sleepiness (Epworth ≥10) despite adequate time in bed → sleep clinic
• Severe evening leg discomfort → ferritin, alpha-2-delta-ligand trial
• Falling asleep at the wheel → urgent OSA / narcolepsy workup
• Snoring + morning headaches + treatment-resistant hypertension → strong OSA association

DACH-specific resources. Germany: DGSM (Deutsche Gesellschaft für Schlafforschung und Schlafmedizin) maintains the list of accredited Schlafmedizinische Zentren. Polysomnography is covered by statutory health insurance (GKV) via the stepwise BUB-Richtlinie. CPAP devices and consumables are reimbursed as Hilfsmittel with ongoing adherence verification. The somnio app is prescribable as DiGA for CBT-I. Austria: ÖGSM, sleep labs at AKH Wien, Graz, Innsbruck. Switzerland: SGSSC, Fähigkeitsausweis Schlafmedizin is the formal subspecialty.

Sleep tracking has grown into a real but limited tool. Polysomnography (PSG, the in-lab gold standard) is the reference, but even between trained human scorers, epoch-by-epoch agreement is only ~83 percent (Cohen's κ ~0.80; Arnal, Sleep, 2020). That ceiling matters. No algorithm can beat the inter-rater limit of PSG itself.

Wrist and ring wearables: TST yes, staging approximate. A 2025 SLEEP Advances validation against PSG (Schyvens, n=62) gives clean numbers on total sleep time mean absolute error:

• Apple Watch Series 8: ~28 min
• Fitbit Charge 5: ~31 min
• Whoop 4.0: ~50 min
• Garmin Vivosmart 4: ~54 min

Sleep-wake sensitivity is uniformly very high (91 to 96 percent). Wake specificity is uniformly poor (29 to 52 percent), so wearables over-detect sleep. If you have fragmented sleep or insomnia, the device probably under-reports it.

Stage classification (Cohen's κ overall vs. PSG):

• Apple Watch S8: ~0.53
• Fitbit Sense: ~0.42
• Whoop 4.0: ~0.37
• Garmin Vivosmart 4: ~0.21

Oura Gen3 with OSSA 2.0 is the best in class at ~0.83 PABAK (Svensson, Sleep Medicine, 2024; 421,045 epochs vs. multi-night ambulatory PSG). REM accuracy ~91 percent. Deep sleep ~76 percent. Note the validation was funded by Oura, so read it with that in mind.

Deep sleep (N3) sensitivity ranges 47 to 70 percent across wrist devices. Oura Gen3 sits at ~80 percent. None of them are measuring SWS directly. They're estimating it from heart rate and motion proxies that line up at the group level but are noisy for any one person.

Home EEG: the only category where you genuinely measure brainwaves. The Dreem 2/3 headband (Arnal, Sleep, 2020) hit N3 κ 0.74, which matches human inter-rater agreement. The consumer device was discontinued in 2021; Beacon Biosignals continues it for research. Muse S Athena ($475) is the current consumer choice. Forehead 4-channel EEG, κ ~0.76 in vendor-supported validations. Z-Machine Insight+ does sleep/wake only.

Closed-loop auditory stimulation (CLAS, pink-noise pulses timed to your slow-wave EEG). These can boost slow-wave activity (~8 percent) and memory consolidation in small studies (Ngo, Neuron, 2013; Papalambros, Frontiers in Human Neuroscience, 2017). The SWA enhancement replicates. The memory benefit is less consistent. No FDA-cleared peer-reviewed consumer CLAS device exists as of May 2026. Philips SmartSleep was discontinued in 2023, Dreem in 2021.

Apnea screening. The Withings Sleep Analyzer (Edouard, Journal of Clinical Sleep Medicine, 2021; n=118) detects moderate-severe sleep apnea with AUROC 0.93 to 0.95, on par with Type III HSAT. A positive result needs HSAT or PSG for diagnosis. Treatment requires it for insurance reimbursement.

HRV during sleep. Overnight RMSSD (the heart-rate-variability metric most reflective of vagal tone) lines up with morning RMSSD but is muddied by sleep stage variability and arousals. Morning HRV is methodologically cleaner. Measure supine for 5 minutes, pre-caffeine, post-void. Track 7-day rolling means, not single days.

Buying guide:

• For TST tracking: any modern wearable. Apple Watch (lowest MAE wrist) or Oura (best ring).
• For SWS measurement: Muse S Athena. Expect 75 to 85 percent accuracy at the stage level.
• For apnea screening: Withings Sleep Analyzer (~€130) at the AHI ≥15 threshold, then a clinic.
• For HRV-based recovery: Polar H10 chest strap + a morning RMSSD app (HRV4Training) is the cheapest gold-standard option (~$90, no subscription).

Vendor-funded validations should be discounted (Whoop, Eight Sleep, Oura Gen4 lack independent peer review). The Schyvens 2025 cohort, the Robbins 2024 Sensors head-to-head, and the Arnal 2020 Dreem paper are the cleanest references.

Most "sleep supplements" target falling asleep, anxiety, or mood, not the depth of slow-wave sleep, which is the longevity-relevant fraction. The gap between marketing and mechanism is the central honesty problem in this category. See our deep sleep guide for the SWS-specific analysis. Here's the broader picture.

What actually has EU-authorised claims for sleep:

• Melatonin has two: "contributes to the alleviation of subjective feelings of jet lag" (≥0.5 mg per portion, close to bedtime on travel day) and "contributes to the reduction of time taken to fall asleep" (1 mg per quantified portion, close to bedtime). That's it. Higher doses are not more effective. A 2013 PLoS ONE meta-analysis (Ferracioli-Oda) showed no dose-response for sleep quality. The 10 mg gummies common in US retail are pharmacologically excessive and tend to leave you groggy in the morning with vivid dreams and no benefit. In Germany, products above 1 mg drift toward medicinal-product classification (BfR 17 September 2024 statement; OLG Koblenz 9 U 1947/22 May 2023 softened the line for ≤1 mg).

What helps onset and subjective quality (not depth):

• L-theanine (an amino acid found in tea that promotes alpha-wave EEG, 200 mg, 30 to 60 min before bed). A 2019 Nutrients RCT (Hidese) showed PSQI total improved over 4 weeks. Mechanism: alpha-wave promotion + anxiolysis. Best for rumination-driven onset insomnia. No EU claim.
• Saffron extract (affron) 14 to 28 mg/day over 4 to 6 weeks improved subjective sleep quality (Lopresti, Journal of Clinical Sleep Medicine, 2020; Pachikian, Nutrients, 2021). The effect is serotonergic / mood-mediated, not SWS-specific. Both trials industry-funded. No EU claim.
• Magnesium bisglycinate 200 to 400 mg elemental in the evening. EFSA UL 250 mg supplemental. A 2021 BMC Complementary Medicine and Therapies meta-analysis (Mah and Pitre) showed SOL reduction ~17 min in deficient/elderly populations. The glycine carrier is part of the story (see deep sleep guide). EU authorised claim: "contributes to a reduction of tiredness and fatigue."
• Glycine 3 g 30 min before bed. A 2007 small RCT (Yamadera) shortened latency to SWS and improved subjective quality, but did not increase the SWS amount. Effect via NMDA-mediated peripheral vasodilation (Kawai, Neuropsychopharmacology, 2015). No EU claim.

What's mostly hype:

• Taurine as a standalone sleep aid: the receptor mechanism (partial GABA-A, glycine receptor) is plausible, but no human RCT with a primary sleep endpoint exists.
• "Sleep formulas" combining 8 to 12 ingredients at sub-therapeutic doses are basically expensive multivitamins.
• Diphenhydramine/doxylamine (OTC anti-histamine sleep aids) carry anticholinergic load (a class effect that impairs cognition and memory, especially in older adults). Chronic use is linked to dementia risk in older adults. AGS Beers criteria recommend against it in adults ≥65.

For deep sleep specifically: the framework. The honest answer is that the SWS levers are mostly behavioural (sleep restriction paradoxically rebounds SWS, exercise raises SWS, a cool room enables it) and substance avoidance (alcohol suppresses first-half SWS, late caffeine cuts SWS). Pharmacological SWS enhancement (gaboxadol/THIP, tiagabine, sodium oxybate) is investigational or limited to specific indications. The supplements with the most defensible SWS framing are magnesium bisglycinate (whose glycine carrier is half the story) and the lemon verbena + zinc combination. Neither has strong polysomnographic human evidence. See the deep sleep guide for the honest breakdown.

The supplement principle. If a product promises to deepen your sleep, ask one question: was that measured with polysomnographic EEG, with a wearable algorithm, or only with a questionnaire? Almost all "deep sleep" claims in the consumer-supplement space are wearable-derived or subjective. The mechanism stories may be real. The SWS-specific human evidence usually isn't.

"I can't fall asleep." Check the basics first: bedroom temperature 16 to 19 °C, last caffeine ≥8 hours ago, no alcohol within 3 hours, no screens or warm lights in the 60 minutes before bed, fixed wake time. If the problem holds ≥3 months, do CBT-I (sleep restriction + stimulus control). If it started within weeks of a new medication, suspect the medication.

"I wake at 3 AM and can't get back to sleep." This is a fragmentation problem, usually one of: alcohol (sleep fragments as blood alcohol falls), apnea (snoring + witnessed apneas + morning headache → STOP-BANG screen → HSAT/PSG), depression or anxiety (the 3 AM awakening is classic for depression), nocturia (cardiometabolic causes), or perimenopausal hot flashes. Treat the underlying cause.

"I sleep 8 hours but wake exhausted." Almost always a quality problem: undiagnosed OSA, poor sleep efficiency, fragmentation from alcohol or late food, depression. Get a wearable to check resting heart rate (high) and HRV (low) overnight as proxies for autonomic load. If snoring is present, screen for apnea. If you have mood symptoms, see a clinician.

"I'm a night owl and can't function in the morning." Probably delayed sleep-wake phase. DSWPD is a diagnosed condition. A sleep clinician (Schlafmediziner) can confirm it and supervise. Standard protocols combine morning outdoor light within 30 minutes of (target) waking with low-dose melatonin (0.3 to 0.5 mg) 5 to 7 hours before target bedtime, shifting the clock 30 to 60 minutes per week. Sustained effort over 2 to 4 weeks. If your chronotype is genuinely late, accept it and adjust your schedule rather than fighting it forever. Roenneberg's social-jetlag data suggest you'll lose either way.

"I get jet lag for a week." Eastward travel is harder than westward because the human circadian system delays more easily than it advances (Khalsa 2003). What works: light timing (morning light at destination for eastward, evening light at destination for westward), low-dose melatonin (0.3 to 0.5 mg) timed to the new dim-light melatonin onset, and short caffeine in the morning. Pre-shifting your schedule 2 to 3 days before travel helps.

"I snore loudly and my partner complains." Screen for OSA: STOP-BANG, and if ≥3 points, push for HSAT or PSG. Snoring alone doesn't equal apnea. But loud snoring + daytime sleepiness + observed apneas is the high-pretest triad.

"I wake up screaming or thrashing." REM behavior disorder (acting out dreams) needs a neurology referral. The 2019 Postuma cohort showed 73.5 percent phenoconversion to PD/DLB/MSA within 12 years. Sleepwalking and night terrors (NREM parasomnias) are usually triggered by sleep deprivation, alcohol, or fever. Treat the trigger, safety-proof the environment.

"My legs feel weird and I can't stop moving them." RLS pattern (worse at rest, evenings, relieved by movement). Check ferritin and transferrin saturation. Ferritin <75 ng/mL → consider IV iron (AASM 2024). Alpha-2-delta ligands (gabapentin enacarbil, pregabalin) first-line; avoid long-term dopamine agonists.

"I take sleeping pills every night and don't want to." Don't quit cold turkey from benzodiazepines or Z-drugs. Rebound insomnia is severe. Taper with medical supervision over 4 to 8+ weeks. Replace with CBT-I, ideally in parallel with the taper rather than after. Daridorexant or another DORA is a safer bridge than a Z-drug if you need pharmacological support.