Sleep is the largest thing this site had never covered, and the hardest of the big items to write about honestly — because the gap between how confident the popular literature sounds and how confident the evidence permits is wider here than anywhere else in this guide.
So this chapter is organised around that gap. What is well established, what is probably true, and what is being sold to you.
1. Duration: the U-shape, and what it does and does not show
Plot mortality against habitual sleep duration and you get a U. Around seven to eight hours sits at the bottom; risk rises on both sides. This is one of the most reproduced findings in epidemiology, across dozens of cohorts and millions of people. Strong as an association.
What it does not show is symmetry of meaning. The short-sleep arm has a plausible mechanism and some experimental support: restrict healthy volunteers' sleep in a laboratory and blood pressure rises, insulin sensitivity falls and appetite regulation shifts within days. The long-sleep arm has almost none. Nobody thinks nine hours of sleep is harmful in itself; long sleep is far more likely to be a symptom of something else — illness, depression, poor sleep quality requiring more time in bed to compensate. Contested
The practical consequence: if you naturally sleep nine hours and feel well, the U-shape is not telling you to sleep less. Trying to compress your sleep to hit a number from a graph is one of the few things in this chapter that could actively make things worse.
2. Regularity may matter more than duration
This is the most interesting recent finding in the field and it points somewhere more useful than the duration advice.
A UK Biobank analysis calculated a Sleep Regularity Index — how similar your sleep and wake timing is from one day to the next — from more than ten million hours of accelerometer data across 60,977 people. Compared with the least regular fifth, the four more regular fifths had a 20–48% lower risk of death from any cause, 16–39% lower cancer mortality and 22–57% lower cardiometabolic mortality, after adjusting for age, sex, ethnicity, sociodemographic, lifestyle and health factors. And regularity predicted mortality more strongly than duration did. Moderate
Why this is worth acting on despite being observational: the advice it generates is close to free, has no plausible downside, and is easier to follow than a duration target. You cannot reliably decide to fall asleep. You can decide when to get up.
Graded Moderate, not Strong, deliberately. It is a single large cohort in an older, healthier-than-average, largely white British population, with one week of accelerometry standing in for a lifetime of habit. Irregular sleep also travels with shift work, illness, caring responsibilities and poverty, and adjustment does not fully remove any of those. It is a good finding. It is not a trial.
3. Sleep apnoea: the part of this chapter with a treatment
Everything else here is about habits. This is a diagnosable condition with an effective treatment, it is substantially under-diagnosed, and it is the reason this chapter sits in a cardiovascular guide.
Obstructive sleep apnoea is repeated collapse of the upper airway during sleep, each episode ending in a brief arousal the sleeper does not remember. The pattern — oxygen dips, surges of sympathetic activity, hundreds of times a night — drives blood pressure up and is associated with atrial fibrillation, stroke and type 2 diabetes.
Worth asking your GP about if several of these apply
- Loud habitual snoring, especially with pauses that someone else has noticed.
- Waking unrefreshed regardless of how long you were in bed.
- Daytime sleepiness that intrudes — dozing off while reading, in meetings, at traffic lights. Sleepiness at the wheel is a reason to raise it urgently.
- Blood pressure that does not settle on three drugs. Resistant hypertension is a classic presentation and one of the commonest missed ones.
- Morning headaches, nocturia, or waking gasping.
- A large collar size, a crowded oropharynx, or a receding jaw — and note that it is not only a condition of large men. It is under-recognised in women, where it more often presents as fatigue and insomnia than as classical snoring.
Diagnosis is a home sleep study in most cases — a device worn overnight rather than a night in a laboratory. Treatment for moderate to severe disease is usually CPAP, which reliably abolishes the events and reliably improves sleepiness and quality of life. Strong
One honest complication. Randomised trials of CPAP — SAVE, RICCADSA and ISAACC among them — have not shown a reduction in major cardiovascular events on intention-to-treat analysis, which sits oddly beside the strong observational association. The most likely explanation is adherence — average nightly use in those trials was low, and a mask worn for three hours leaves most of the night untreated — but it has not been proved, and it should be stated rather than skated over. Contested The case for treating moderate to severe apnoea rests securely on symptoms, function and driving safety; the cardiovascular case is weaker than the mechanism suggests it ought to be.
4. Insomnia: the treatment is not a tablet
For persistent insomnia, cognitive behavioural therapy for insomnia is the recommended first-line treatment ahead of medication, and the evidence for it is good. Strong It works through sleep restriction, stimulus control and unpicking the beliefs that keep people awake, and its effects outlast the course in a way that sleeping tablets' do not.
It is also unglamorous, initially unpleasant — sleep restriction makes things worse for a week or two before it works — and harder to access than a prescription. In England it is available through digital programmes on the NHS in many areas, and a GP can advise on local routes.
Hypnotics have a role that is genuinely narrow: short courses, for short crises. Beyond a few weeks the benefit shrinks, tolerance builds, and in older adults the falls risk is real.
5. What the evidence does and does not support
| Claim | Where the evidence sits |
|---|---|
| Consistent sleep and wake times | Best value-for-effort item in the chapter. Observational but consistent, and free. |
| Treating diagnosed sleep apnoea | Strong for symptoms and function; contested for cardiovascular events. |
| CBT for insomnia over hypnotics | Strong, and the recommended first line. |
| A dark, cool, quiet bedroom | Reasonable, cheap, modest. Nobody has run the trial and nobody needs to. |
| Caffeine cut-off in the afternoon | Real mechanism — caffeine's half-life is around five hours, longer in some people. Individual sensitivity varies enormously. |
| Alcohol as a sleep aid | Reliably counterproductive. It shortens time to sleep and degrades the second half of the night. |
| Melatonin for ordinary insomnia | Weak. It is a circadian signal, not a sedative, and is most useful for jet lag and shift work. Prescription-only in the UK, unlike the US. |
| Sleep-tracking wearables | Poor at staging sleep, decent at timing and regularity — which is the thing that matters. Use them for consistency, not for a nightly score. |
| Most sleep supplements | No meaningful human outcome evidence. See the supplements page for how this site grades that category. |
What to actually do
Fix the wake time first
The same time every day, including weekends, within about an hour. It is the single lever that most improves regularity, and unlike bedtime it is under your control.
Get apnoea excluded if any of the flags apply
Especially if your blood pressure is stubborn or you are sleepy at the wheel. This is the one item here that is a medical problem with a medical answer.
If insomnia has lasted more than a few weeks, ask about CBT-I
Not a tablet, and not another article about sleep hygiene — sleep hygiene advice alone is a weak treatment for established insomnia.
Stop optimising
The returns beyond regular timing, a decent environment and excluding a treatable disorder are small, and the anxiety generated by pursuing them is not.