How much sleep do you need?
About 7 to 8 hours a night. Across nearly 1.4 million adults, sleeping much less — or much more — than that was linked to a higher risk of dying early, in a U-shaped pattern. But the two ends mean different things, and "more sleep is risky" does not mean you should sleep less.
What the evidence supports
- ✓ ~7–8 h is linked to the lowest mortality across ~1.4 million adults1
- ✓ Short sleep (<7 h, esp. ≤5 h) linked to ~12% higher mortality — with direct metabolic harm in experiments
- ✓ A consistent, U-shaped pattern with no publication bias
What it doesn't mean
- ✗ That long sleep causes harm — it's usually a marker of illness, not a reason to sleep less
- ✗ Proof of cause — the data are observational and self-reported
- ✗ A precise personal number — 6–8 h shows no evident harm, and needs vary
Where this comes from
- Why the middle is best
- The two ends of the U are not the same. Too little sleep does measurable harm: controlled experiments show that cutting sleep worsens glucose control and raises appetite hormones within days, which plausibly feeds long-term risk. Too much sleep is different — regularly sleeping nine or ten hours tends to come with underlying illness, so long sleep is best read as a signal that something is wrong, not as a cause of early death. That distinction is the whole reason the advice is to protect your sleep, not to cut it.
- What the human evidence shows
- The anchor is a confirmed meta-analysis of 16 studies and 27 cohorts — 1,382,999 adults across eight countries, with 112,566 deaths over 4 to 25 years.1 Compared with a 7–8 hour reference, short sleep (under 7 hours, often 5 or fewer) carried about 12% higher all-cause mortality (relative risk 1.12), and long sleep (over 8–9 hours) about 30% higher (relative risk 1.30). The authors note that 6 to 8 hours shows no evident harm, while consistently sleeping 5 hours or less marks out a higher-risk group.
- Where the evidence stops
- This is an association from observational cohorts, with sleep reported by the participants themselves — so it shows a link, not proof of cause, and we say "associated with" rather than "causes". The practical, honest message is narrow and firm: protect a consistent 7–8 hours. If you routinely get 5 hours or less, that is the pattern worth changing; if you find you genuinely need 9 or 10, treat it as a reason to get checked, not a habit to trim.
Why "long sleep is risky" is not advice to sleep less: the higher risk at the long end is almost certainly because illness makes people sleep more — the sickness comes first. Cutting your sleep to chase a lower number would remove something protective and change nothing about the underlying cause. The only actionable takeaway is to guard your 7–8 hours. We sell nothing and recommend no sleep product.
At a glance
Sources
- Sleep duration and all-cause mortality: a systematic review and meta-analysis of prospective studies. Sleep, 2010. PMC2864873
Bottom line for you
Aim for a consistent 7 to 8 hours. That range is linked to the lowest mortality across well over a million adults, and there's no evident harm anywhere from 6 to 8. If you regularly run on 5 hours or less, that is the single most worthwhile thing to change.
Don't misread the U-shape. The higher risk among long sleepers reflects illness that causes the extra sleep, not sleep that causes illness — so there is no version of this evidence that tells you to sleep less. If you find you truly need 9 or 10 hours night after night, that's worth a conversation with your clinician, not a target to cut towards.
Get the evidence, not the hype
We'll send you the full Evidence Index table — every food and method we've rated by the strength of the human evidence, all in one place. Human studies only, and no supplements.
Educational information based on peer-reviewed human studies. Not medical advice — persistent sleep problems or a need for very long sleep are worth discussing with a clinician. AntiAgingGreens never sells or recommends supplements. We report what the evidence shows — and where it stops.