What a night is made of
Sleep runs in cycles of about 90 minutes, four to six of them a night, each moving through light sleep (N1 and N2), deep slow-wave sleep (N3) and REM. The proportions shift through the night: deep sleep dominates the first half, REM the second. That has a practical consequence — going to bed late costs deep sleep, and getting up early costs REM, and both losses are invisible to someone who "slept six hours and feels fine".
| Stage | Share of the night | What it does | What damages it |
|---|---|---|---|
| N1–N2 light | ~50% | Transition and memory consolidation; sleep spindles in N2. | Noise, pain, an uncomfortable temperature. |
| N3 deep (slow-wave) | 15–25%, falling with age | Growth-hormone pulse, tissue repair, glymphatic clearance of the brain, glucose regulation. | Alcohol, late meals, a warm bedroom, late training, apnoea, benzodiazepines. |
| REM | 20–25% | Emotional processing, procedural memory, creativity. | Alcohol above all; cannabis; many antidepressants; short nights. |
Deep sleep declines markedly with age — a 60-year-old may get half the slow-wave sleep of a 25-year-old — and that decline is one of the plausible links between poor sleep and dementia: the glymphatic system that clears amyloid from the brain runs mainly during N3.
The evidence on duration
The relationship between sleep length and mortality is U-shaped, and it has been U-shaped in every large dataset. The Cappuccio 2010 meta-analysis in Sleep pooled 16 cohorts and 1.38 million people: sleeping under about six hours was associated with a 12% higher risk of death, and over eight or nine hours with a 30% higher risk. The long-sleep arm is partly reverse causation — illness makes people sleep more — but the short-sleep arm has held up in studies that adjusted for everything they could.
In UK Biobank (Li et al. 2022, Nature Aging; nearly 500,000 adults aged 38–73), seven hours was the duration associated with the best cognitive performance, mental health and brain structure, with a steady decline either side. A separate UK Biobank analysis (Fan 2020, European Heart Journal; 385,292 people) built a healthy-sleep score — 7–8 hours, morning chronotype, no insomnia, no snoring, no daytime sleepiness — and found the best scorers had 35% lower cardiovascular disease than the worst, independent of genetic risk. Irregularity matters too: a 2024 analysis of 72,000 Biobank wearables found that an irregular schedule was associated with higher mortality even at a normal average duration.
These are cohorts, so the usual caution applies; but the experimental data point the same way. Restricting healthy adults to four or five hours for a week reliably produces insulin resistance comparable to pre-diabetes, raises evening cortisol and ghrelin, and lowers testosterone by 10–15% in young men (Leproult 2011, JAMA).
Sleep apnoea: the one to rule out first
Obstructive sleep apnoea — the airway collapsing repeatedly during sleep — affects an estimated 936 million adults worldwide (Benjafield 2019, Lancet Respiratory Medicine), most of them undiagnosed. In men over forty with a neck over 43 cm or a BMI over 30, it is the most likely explanation for waking unrefreshed, afternoon exhaustion, high blood pressure that will not settle and low morning testosterone. Untreated moderate-to-severe apnoea is associated with roughly double the cardiovascular mortality.
Screening is a questionnaire (STOP-Bang: snoring, tiredness, observed pauses, pressure, BMI, age, neck, gender — three or more is a referral) and diagnosis is a home sleep study. Treatment is a physician's decision — CPAP, a mandibular device, weight loss — and it is the one intervention on this page that can change a life in a week. No protocol we write for a man over 45 who snores starts without the question being asked.
What actually helps
| Lever | Evidence | How |
|---|---|---|
| Morning light | Strong — the circadian anchor | 10–30 minutes outdoors within an hour of waking, no sunglasses. On a dark morning, a 10,000-lux lamp. |
| A fixed wake time | Strong — regularity data above | Same time every day including weekends; the bedtime follows. |
| Cool bedroom | Moderate | 18–19 °C. Core temperature has to fall to initiate sleep. |
| Warm bath or shower 1–2 h before bed | Moderate — meta-analysis of 17 studies (Haghayegh 2019) | Paradoxically speeds the drop in core temperature; fell asleep ~10 minutes faster. |
| Caffeine cut-off | Strong | Half-life 5–6 hours; nothing after about 2 p.m. A 400 mg dose six hours before bed cut sleep by an hour in a controlled study. |
| Alcohol | Strong (negative) | Sedates, then fragments the second half and suppresses REM. Two drinks cost measurable sleep quality; none is the sleep answer. |
| Last meal 3 h before bed | Moderate | Late meals raise core temperature and glucose overnight. |
| Dim evening light | Moderate | Lights low after sunset; screens are less of the problem than the content and the brightness. |
| Training | Strong | Regular aerobic and resistance training deepens N3. Hard sessions finished more than three hours before bed. |
| CBT-I | Strong — first-line for insomnia in every guideline | Six to eight sessions, in person or app-based. Outperforms sleeping pills at a year. |
| Magnesium, glycine | Early, small trials | Magnesium glycinate 200–400 mg or glycine 3 g; modest effects, low risk. |
| Sleeping pills | Negative for architecture | Z-drugs and benzodiazepines reduce deep sleep and are associated with higher mortality in cohorts. Short-term, physician-managed only. |
What to measure
- Duration, regularity and timing from a wearable or a two-week diary. Average and standard deviation of wake time; aim for a wake-time spread under 30 minutes.
- Deep and REM estimates from the wearable — imprecise night to night but useful as a trend, and sensitive to alcohol in a way that is instructive.
- Resting heart rate and HRV overnight: the earliest signal that alcohol, a late meal or over-training cost you.
- STOP-Bang once, and a home sleep study if it scores three or more.
- Labs: morning testosterone, HbA1c, fasting insulin, hsCRP, ferritin — all of which move with sleep — and iron in anyone with restless legs.
Where peptides and therapies fit
The sleep goal page is the protocol built around this article: the cheap levers above come first and are not optional, and the compound layer is added only when they are in place. The compound most asked about is DSIP, delta sleep-inducing peptide, which is studied for sleep architecture and has small, old human studies; it is not a sedative and does nothing for a schedule that is wrong. Among therapies, sauna in the early evening is one of the better-evidenced ways to deepen slow-wave sleep through the temperature mechanism. And for anyone whose problem is waking at 3 a.m. with a racing mind, hormones after 40 and the cortisol question come before anything else.
Common mistakes
- Fixing bedtime instead of wake time. The wake time sets the clock; the bedtime is a result.
- A nightcap. It is the most common and most damaging sleep habit in people over forty.
- Catching up at the weekend. Two late mornings produce a Monday that feels like jet lag and the irregularity itself carries risk.
- Ignoring the snoring. Apnoea is treatable and, untreated, undoes every other lever on this page.
- Trusting a wearable's deep-sleep number night to night. Use the monthly trend.
- Reaching for a compound before the room is cool, dark and the coffee has stopped.
Frequently asked questions
How many hours of sleep are best for longevity?
Seven to eight. In the 1.38-million-person Cappuccio meta-analysis both short (under six) and long (over nine) sleep were associated with higher mortality, and in UK Biobank seven hours was associated with the best cognition and mental health. Regularity matters almost as much as the number.
Is sleeping too much really harmful?
The long-sleep arm of the U-curve is partly reverse causation — illness and depression make people sleep more — so it is less certain than the short-sleep arm. If you consistently need more than nine hours, that is worth investigating rather than fixing with an alarm.
How do I know if I have sleep apnoea?
Snoring, witnessed pauses, waking unrefreshed, afternoon sleepiness and resistant high blood pressure are the signs. The STOP-Bang questionnaire screens it; three or more points is a referral for a home sleep study. It is very common and very under-diagnosed in men over forty.
Do sleep supplements work?
Modestly, at best. Magnesium and glycine have small positive trials; melatonin helps timing more than depth; most "sleep stacks" have no evidence. None of them beats a fixed wake time, morning light, a cool room and no alcohol.
Can a peptide improve sleep?
DSIP is studied for sleep architecture and has small human studies; it is not a sedative. The sleep protocol we write puts it, if at all, on top of the behavioural levers, after an apnoea screen and a look at the labs that sleep loss moves.
Want a plan built on your own numbers?
The sleep protocol is written from your sleep data, your labs and your evenings — the cheap fixes first, the compound layer only where it earns a place. €399, follow-up included.
Build my protocol Book a 15-minute call