Sleep
Four things about sleep independently predict how long you live: how much, how regular, how dark, and whether you stop breathing at night. Most of what gets sold to fix sleep — pills, supplements, trackers chasing a deep-sleep score — addresses none of them.
Sleep 7–8 hours on a regular schedule, anchored by a fixed wake time; keep the bedroom dark and cool; get screened for sleep apnea if you snore; and treat chronic insomnia with CBT-I before any pill. In large cohorts, short sleep (under about seven hours), long sleep (over about nine), irregular timing and untreated sleep apnea each independently predict death, cardiovascular disease (CVD), dementia, type 2 diabetes and depression. The American Heart Association added sleep to its core cardiovascular-health metric, Life's Essential 8, in 2022.[1] Most of that evidence comes from large, consistent observational studies, so it shows associations; where randomised trials exist, the sections below say so.
Why sleep matters for aging
Sleep is when growth hormone is released, cortisol bottoms out and autonomic tone resets; bad sleep plausibly speeds the processes behind late-life disease, and the evidence that it does is stronger than the evidence that fixing it winds them back.
- The glymphatic system — the brain's proposed waste drainage — is said to clear amyloid-β, the protein that clumps in Alzheimer's, mainly during deep sleep. That is contested: a 2024 mouse study found clearance lower during sleep,[2] while the original group still finds it driven by sleep.[3] In people, an industry-funded 2026 trial found no difference in overnight blood amyloid-β or tau between a night's sleep and a night awake.[4] A Framingham follow-up linked each percentage-point yearly fall in deep sleep to about 27% higher dementia risk, but early dementia may itself erode deep sleep.[5]
- Hormones. In men, roughly 70% of daily growth hormone comes in early sleep.[6] Restricting ten healthy young men to five hours in bed for a week cut daytime testosterone by 10–15%, against a normal decline of 1–2% a year.[7] See Testosterone therapy.
- Epigenetic aging. Among 3,795 older adults, short sleepers ran about 1.3 years older than their age on one DNA-based "epigenetic clock", in a cross-sectional analysis.[8] In a secondary analysis of a CBT-I trial in 231 adults over 60, a cellular-senescence marker rose over two years in the control group and held flat under CBT-I — suggestive, not settled.[9]
How much: seven to eight hours, and don't chase deep sleep
Against a reference of 7–8 hours a night, under seven hours is associated with about 14% higher mortality and nine or more with about 34%.[10] The long-sleep figure partly reflects illness — depression, untreated apnea, frailty — rather than the sleep itself.
A night runs four to six cycles of about 90 minutes through light sleep, deep (slow-wave) sleep and rapid-eye-movement (REM) sleep; deep sleep comes early, REM late. The brain defends deep sleep: across a week of 6-hour nights REM absorbed the loss, and after 40 hours awake deep sleep rebounded to 140–152% of baseline.[11] The one everyday thing that seems to raise it is alcohol, which probably adds deep sleep early in the night — a disputed finding[12][13] — while cutting REM from about two standard drinks; falling asleep faster showed up only at about five.[14] That is why a tracker may flatter you after a nightcap.
Consumer sleep trackers mis-stage deep sleep anyway. Against a sleep-lab recording, current wrist devices ran from 25 minutes under to 44 minutes over on deep sleep by brand, and one that lumps deep and REM together was 73 minutes over; these are group averages, not one person wearing two devices.[15] "Orthosomnia" — anxiously chasing tracker scores — was named in a report of three patients whose pursuit may have worsened their insomnia; whether trackers cause it is barely tested, and the one small controlled comparison found no harm.[16][17] Sleep architecture covers the stages, how they change with age, and why deep sleep is a poor thing to chase.
When you sleep matters as much as how long
Sleep regularity predicts death at least as strongly as duration. In about 61,000 UK adults wearing wrist trackers, the most regular fifth had 48% lower mortality than the least regular, falling to about 30% after fuller adjustment, and 55% lower cardiometabolic and 39% lower cancer mortality before that adjustment; two senior authors have lighting-industry ties.[18] Irregular sleepers also had 26% more major cardiovascular events, and enough total hours did not offset it.[19] A 2025 review judged irregularity an independent risk factor for cardiometabolic disease,[20] and irregular timing predicted dementia over a decade of follow-up.[21] All of this is observational.
Clock time matters too. In 88,026 adults, cardiovascular risk was lowest for falling asleep between 10 and 11 pm — about 25% higher at midnight or later, about 24% higher before 10 pm — independent of sleep length.[22]
A master clock in the brain keeps the cellular clocks in every tissue in step; it runs slightly longer than 24 hours (about 24.2) and is reset daily by light, meal timing and routine.[23] Weekend catch-up sleep restores some alertness but not the weekday metabolic cost. Night-shift work, which fights that clock directly, is rated probably carcinogenic to humans (Group 2A) by the World Health Organization's cancer agency, on limited human evidence for breast, prostate and colorectal cancer and sufficient evidence in animals.[24] Circadian rhythms covers light, meal timing, jet lag and shift work.
Bright days, dark and cool nights
Light is the strongest setter of the body clock, and evening light does more than most people assume: five evening hours of light dimmer than most living rooms (about 25 lux) halved melatonin in the average young adult.[25]
Light at night also tracks with health. In about 89,000 UK adults wearing light sensors, the brightest-night group had 21–34% higher mortality than the darkest,[26] and 53% higher risk of type 2 diabetes, independent of genetic risk.[27] That is one cohort and one research group, so confounding remains live; satellite-measured outdoor light lost its link with death once air pollution and noise were accounted for.
A cool bedroom helps, though the often-quoted 18–20 °C is convention: the one home-monitoring study in older adults found sleep most efficient at 20–25 °C, with wide individual variation.[28]
Sleep apnea: the underdiagnosed cardiovascular risk
Obstructive sleep apnea (OSA) is the repeated collapse of the upper airway during sleep, graded by the apnea-hypopnea index — blockages per hour. An estimated 936 million adults aged 30–69 have it, most undiagnosed — a ResMed-funded figure, so trust its direction more than its precision.[29]
The mortality signal sits in the severe grade. Pooling 27 cohorts and over 3 million people, severe OSA carried roughly double the all-cause mortality (hazard ratio 2.13) and nearly triple the cardiovascular mortality; for mild OSA (1.19) and moderate (1.28) the uncertainty ranges crossed 1, compatible with no effect.[30] In about 1,000 patients, apnea was linked to roughly double the risk of stroke or death.[31] The case for treating severe OSA on mortality grounds is strong; for mild OSA, symptoms are the reason to treat.
Screening is cheap. The STOP-Bang questionnaire — eight yes/no questions on snoring, tiredness, observed apneas, high blood pressure, body mass index, age, neck size and sex — flags risk; three or more "yes" answers warrant a sleep study.[32] A home sleep test is now first-line for most adults.
CPAP (continuous positive airway pressure) works when used: at four or more hours on at least 70% of nights it makes people less sleepy and lowers blood pressure, substantially so when blood pressure resists medication.[33] Whether it prevents heart attacks and strokes is unresolved: the SAVE trial in 2,717 adults found no reduction over 3.7 years, though average use was only 3.3 hours a night, and the analyses suggesting benefit above four hours are not randomised. Treating apnea after atrial-fibrillation ablation did not cut recurrence (57% in both arms).[34]
Weight loss helps obesity-driven OSA: a 10% weight loss predicted a 26% fall in the apnea-hypopnea index.[35] Tirzepatide cut it by about 20 events an hour without CPAP, and 24 with it, over 52 weeks in adults with moderate-to-severe OSA and obesity.[36] See Ozempic-class drugs. Sleep-disordered breathing covers diagnosis and the alternatives to CPAP, including mouth taping and myofunctional therapy.
Chronic insomnia: therapy before pills
Roughly 10–15% of adults have chronic insomnia — trouble falling or staying asleep at least three nights a week for three months. Cognitive behavioural therapy for insomnia (CBT-I) comes first, and it mostly keeps working after it ends,[37] while sleeping pills stop working when you stop them. Its core includes stimulus control and sleep restriction. Digital CBT-I clearly works and is the right move without a therapist, but the one trial against in-person therapy found it worse, so it is not a proven equal.
When a pill is genuinely needed — as a bridge, for a short crisis, or after CBT-I has failed — the usual choice is an orexin blocker (suvorexant, lemborexant, daridorexant), which damps the brain's wakefulness signal rather than sedating you. They help by small margins, with no tolerance, withdrawal or rebound insomnia in trials of up to six months.[38] They are still US controlled substances,[39] and across 11 trials and 7,703 patients they roughly tripled sleep paralysis and doubled excessive daytime sleepiness — uncommon events.[40] The preference rests on fewer known harms, not proven safety.
The older options carry more baggage when used long term: falls, fractures and cognitive impairment. Z-drugs (zolpidem and relatives) are for short courses of four weeks or less; pooled observational data put hip-fracture risk 90% higher with Z-drugs and 52% higher with benzodiazepines, more than double in the first weeks of a new prescription.[41] The dementia signal belongs to anticholinergics such as the over-the-counter antihistamines (diphenhydramine — Benadryl, ZzzQuil), and is observational: early dementia may drive both the poor sleep and the drug use.[42] The Beers Criteria say to avoid benzodiazepines and first-generation antihistamines from age 65; below that there is no evidence-based cut-off, but the antihistamines still work poorly.[43] A 2025 model projected that if Americans over 50 stopped prescription sleep medication, lifetime falls would drop 8.5% and life expectancy rise by about 40 days — a model built on observational risks.[44] Treating chronic insomnia covers the full protocol and special cases.
Naps: short helps, long is a warning sign
A short nap is one of the few sleep habits backed by randomised trials. Pooling 22 of them, all in physically active young men, napping gave a moderate-to-large boost to thinking performance and improved physical performance for the next few hours, largest with naps of 30 to under 60 minutes.[45] A nap under 30 minutes is a quick alertness boost with little grogginess; memory gains appeared at about 30 minutes.
Habitual long naps point the other way. A 2026 umbrella review of 16 meta-analyses found naps over 60 minutes associated with roughly 30% higher coronary-heart-disease risk and higher diabetes, obesity and all-cause mortality risk.[46] In older adults measured with wrist devices, each extra hour of daily napping went with about 13% higher mortality.[47] A genetic study found about 52% higher odds of coronary artery disease with napping, but its design biases toward the observed link and it captures how often people nap, not how long.[48]
The harm is conditional, and mostly a marker of underlying illness rather than its cause. Across 44 cohorts and more than 1.8 million people it was confined to naps of 30 minutes or longer, and napping went with lower risk of cognitive impairment and muscle loss.[49] In 116,632 adults across 21 countries, the excess appeared only in people sleeping more than six hours a night.[50] In older adults, heavy napping is more likely a sign of decline than its cause: nappers' slightly higher C-reactive protein, an inflammation marker, all but vanished after adjusting for health and lifestyle.[51]
Daytime naps covers the dose-response, sleep inertia and the siesta.
Sleep supplements: modest at best
Most "natural" sleep aids give small effects with few side effects; none replaces CBT-I, and none causes meaningful tolerance or withdrawal — which is not the same as harmless for years.
Magnesium has a small, real effect, largest in people with low dietary intake, on a trial base rated low-certainty; the best trial, of magnesium glycinate, found about one extra meaningful responder per seven people treated.[52] Keep supplemental magnesium near 250 mg elemental — the trial dose and the EU upper limit for supplements. L-theanine and glycine are in the same class: small effects, excellent safety. Ashwagandha (300–600 mg standardised root extract) has low-certainty evidence of a modest effect on stress-driven sleep disruption, and carries rare liver injury. CBD (cannabidiol) interferes with the liver enzymes that clear many prescriptions, including statins and warfarin.
Melatonin works when dosed to the purpose — jet lag and delayed sleep timing: 0.5–5 mg of immediate-release melatonin for jet lag; 0.3–1 mg to shift the clock; slow-release 2 mg did relatively poorly for jet lag.[53] For insomnia the American Academy of Sleep Medicine advises, conditionally, against it.[54] Europe licenses a prolonged-release 2 mg prescription form for insomnia from age 55, for up to three months.[55] That licence, not a melatonin shortage with age, is the case for older adults; see Melatonin. Going above about 5 mg adds nothing and 10 mg is never justified; even 3 mg can spill into the next morning. US products ranged from 83% below to 478% above their label.[56] A preliminary 2025 conference report linked melatonin prescribed for insomnia for a year or more to roughly double the rates of heart failure and death.[57] It drew on prescription records, so it describes people with insomnia severe enough to be medicated, but it is reason to use melatonin strategically, not nightly for years.
Sleep supplements ranks each one.
Practical sleep checklist
- Anchor your wake time. Same time every day, weekends included; let bedtime follow.
- Get morning light: 10–30 minutes outdoors within an hour of waking. Midday daylight runs 50,000–100,000 lux against 500–1,000 indoors.
- Dim the evening. Cut overhead light and screens for the last three hours before bed. A 2022 expert consensus set targets in melanopic lux (light weighted to the receptors that drive the body clock): at least 250 by day, under 10 in the three hours before bed, under 1 while asleep.[58]
- Aim to be asleep roughly between 10 and 11:30 pm where your schedule allows.
- Cool, dark bedroom: blackout curtains, no glowing electronics. A warm bath or shower (40–42.5 °C) one to two hours before bed speeds the core-temperature drop and shortens the time to fall asleep.[59]
- Size the caffeine cut-off to the dose. One cup (about 100 mg) had no measurable effect even four hours before bed; 400 mg altered sleep within 12 hours of bedtime.[60] See Coffee.
- Move the last drink as early as you can, or skip it on nights when sleep matters. Three hours before bed is only a rule of thumb: a moderate dose six hours before bed, with breath alcohol back to zero by bedtime, still cut total sleep and REM sleep and doubled waking in the second half of the night.[61] See Alcohol.
- Eat earlier: finish your last meal 2–3 hours before bed, mainly to spare yourself night-time reflux; a later supper hasn't been shown to worsen sleep itself. See Fasting.
- Screen yourself for OSA if you snore loudly, have witnessed apneas, high blood pressure, atrial fibrillation or a body mass index of 30 or more: take STOP-Bang and ask for a sleep study on three or more "yes" answers.
- For chronic insomnia, start with CBT-I, digital if no therapist is available.
- Cap naps at 20–30 minutes, early-to-mid afternoon, and give yourself 10–20 minutes before driving. If you regularly need more, find out why.
- Don't tape your mouth without a clinical check of your nose and a screen for apnea.
What's overrated
- Weekend catch-up sleep. It doesn't repair the weekdays, though sleeping in isn't itself harmful.
- Blue-light-blocking glasses as the sole fix. Evidence is very low certainty; dimming the room matters more.[62]
- High-dose melatonin (10 mg). Nothing above about 5 mg works better.
- Chasing a deep-sleep score. The brain defends deep sleep, trackers misread it, and the one everyday thing that seems to raise it is alcohol.
- Mouth taping as a biohack. Defensible only after an ear-nose-throat check has cleared nasal obstruction.
- Apps and watches in place of a sleep study. Breathing-sound analysis reaches about 90% sensitivity in research settings, but a watch's apnea alert is tuned to be almost never wrong when it fires, at the cost of catching only about 43% of moderate apnea. A silent watch is not a negative test.[63]
- "Don't exercise in the evening." Mostly a myth: a meta-analysis found evening exercise does not disrupt sleep overall, with only hard sessions ending shortly before bed slightly trimming REM.[64] Vigorous exercise within about an hour of bed may still delay sleep in sensitive people.
For the full ordered action list drawn from every pillar, see the healthspan long list.