Sleep and longevity

— What the evidence shows

A meta-analysis of 16 prospective studies (1,382,999 participants, 112,566 deaths) found that short sleepers had a 12% higher risk of death (relative risk 1.12) and long sleepers a 30% higher risk (1.30) than people sleeping in the reference range. Later pooled analyses report the lowest risk at around seven hours.

Most of this evidence relies on self-reported habitual sleep duration recorded once, which is a rough measure of a behaviour that varies night to night and across life.

Sleep quality, disorders such as sleep apnoea, and shift work are separate exposures that are not captured by duration alone.

— Why the U-shaped curve is difficult to interpret

Pooled cohorts covering millions of participants report higher mortality among people reporting short sleep and among people reporting long sleep, with the lowest observed risk usually around seven hours. Describing that pattern is straightforward; reading a cause into either side of it is not.

• The exposure is usually habitual sleep duration reported once — a rough measure of something that varies night to night and across a life.
• Long sleep is plausibly a symptom of existing illness rather than a cause of death, which would produce this pattern without long sleep being harmful.
• Duration says nothing about sleep quality, about disorders such as sleep apnoea, or about shift work — separate exposures with their own associations.
• No trial has randomised people to different sleep durations for decades, so the lowest point of a population curve is not a personal prescription.

— What we know

• The U-shaped association with mortality is reproduced across many cohorts and regions.
• Untreated sleep disorders are associated with cardiovascular and metabolic disease.

— What remains uncertain

• Long sleep is plausibly a symptom of existing illness, depression or low physical activity rather than a cause of death (reverse causation and confounding), which would produce the observed pattern without long sleep being harmful.
• Duration is usually self-reported once, by questionnaire; it is not the same as sleep quality, and the studies say little about quality.
• Studies differ in their reference category and in what they adjust for, which is one reason results vary between cohorts.
• No trial has randomised people to different sleep durations for decades, so causal effects on lifespan are unestablished.
• An optimal individual sleep duration cannot be read off a population curve.

Sources

• Cappuccio et al., Sleep duration and all-cause mortality: a systematic review and meta-analysis, Sleep (2010)
• Yin et al., Relationship of sleep duration with all-cause mortality and cardiovascular events, JAHA (2017)
• Watson et al., Recommended amount of sleep for a healthy adult: a joint consensus statement of the AASM and SRS, Sleep (2015)

Continue exploring

• Physical activityA factor where the same reverse-causation problem appears.
• What affects longevityAll six factors, and the kinds of evidence behind each one.
• What is longevity?The terms behind lifespan, life expectancy and healthspan.
• Can humans actually extend lifespan?Why observational curves cannot be read as instructions.