Does sleep affect how long you live? Both too little and too much track with dying sooner

The evidence that sleep is tied to how long you live is unusually large and unusually consistent: pooled across more than a million people, both short and long sleep predict earlier death, with the lowest risk sitting around seven hours a night [s1][s2]. The catch is that almost all of this is observational, and the most likely explanation for part of the curve — that illness disrupts sleep, rather than sleep disrupting health — is the one a spreadsheet cannot rule out. No trial has shown that changing how long you sleep extends life.

What the data show

The foundational analysis is a 2010 systematic review that pooled 16 studies providing 27 independent cohort samples — 1,382,999 participants followed for 4 to 25 years, with 112,566 deaths recorded [s1]. Compared with a middle reference, short sleep carried a relative risk of death of 1.12 (95% CI, 1.06 to 1.18), and long sleep a relative risk of 1.30 (1.22 to 1.38) [s1]. The striking detail is that long sleep was the stronger signal of the two.

A larger 2017 dose-response meta-analysis sharpened the shape. It found U-shaped associations between sleep duration and all-cause mortality, total cardiovascular disease, coronary heart disease and stroke, with the lowest risk at roughly seven hours a day [s2]. For death from any cause, each hour below seven raised the pooled relative risk by 1.06 (95% CI, 1.04 to 1.07), and each hour above seven raised it by 1.13 (1.11 to 1.15) [s2]. Again the long end of the curve was steeper. For stroke the asymmetry was sharpest: 1.05 per hour of short sleep against 1.18 per hour of long sleep [s2].

The reverse-causation problem

A U-shaped curve is exactly the pattern you would expect if sickness were doing the driving. People in the early, undiagnosed stages of heart failure, cancer, depression or infection often sleep much more than usual; sleep can lengthen for years before a diagnosis. If long sleep is frequently a symptom of illness rather than a cause of it, the steep long-sleep arm of the curve partly measures people who were already dying, not people whom sleeping harmed. This is why the long-sleep signal, though real in the data, is the least trustworthy part of the story.

Short sleep has a more plausible biological path — chronic short sleep is linked to higher blood pressure, impaired glucose regulation and inflammation — but it, too, travels with shift work, poverty, pain and untreated sleep disorders, each an independent risk to health [s2]. The cohorts adjust for what they measure, and self-reported sleep duration is itself a blunt instrument that correlates only loosely with sleep actually recorded. So the honest reading is that both tails of the curve mix cause and consequence in proportions the observational design cannot separate.

What it means for a reader

The practical takeaway is narrower than the headline number suggests. The data are strong enough to say that habitual sleep far outside the roughly seven-hour range is a marker worth taking seriously — not because a clock reading is destiny, but because persistent very short or very long sleep is often a signal of something else: an untreated disorder, a mood problem, a developing illness [s1][s2]. The useful response to consistently disrupted or excessive sleep is to ask why, not to force a number.

What the evidence does not support is the inverse of the observation — that a healthy person can extend their life by engineering their sleep to a target. There is no trial in which assigning people to sleep more, or less, changed how long they lived, and it is not obvious such a trial could even be run. The U-shaped curve is one of epidemiology's most reproducible findings and one of its most over-interpreted: it describes a population association shot through with reverse causation, not a dial that individuals can turn to buy time.

The claim best supported by the data is the modest one. Sleep that is regularly and markedly abnormal in length is associated with worse survival and deserves attention as a symptom; chasing a precise nightly figure in the name of longevity is a step the evidence does not license.