Sleep duration and mortality

The largest pooled analysis of habitual sleep duration and all-cause mortality combined 16 prospective studies (27 separate study populations) with 1,382,999 participants, 112,566 deaths and follow-up of 4 to 25 years. Sleep was reported by questionnaire. Compared with people who slept about 7 hours, short sleepers had a relative risk (the risk in one group divided by the risk in another) of 1.12 (95% CI 1.06 to 1.18) and long sleepers 1.30 (1.22 to 1.38) [1]. A companion analysis of 15 studies and 474,684 participants found that short sleep was associated with a 48% higher risk of developing or dying of coronary heart disease and 15% higher risk of stroke, and long sleep with 38% and 65% higher risks respectively [6].

In an Australian study of 231,048 middle-aged and older adults followed for 6 years, sleeping less than 7 or more than 9 hours a day was one of six lifestyle risk factors examined. As a single risk factor, short sleep was only marginally associated with mortality (hazard ratio 1.09, where 1.00 means no difference), while long sleep was associated with 44% higher mortality. Long sleep combined with physical inactivity was associated with 2.68 times the mortality of having no lifestyle risk factor [7].

Sleep regularity has been measured with the Sleep Regularity Index, which compares the sleep-wake state at each time of day with the same time on the next day. In 60,977 UK Biobank participants with a mean age of about 63, more than 10 million hours of wrist accelerometer data were used to calculate the index, and 1,859 deaths occurred over a mean of 6.3 years. Compared with the least regular fifth, the other four fifths had 20 to 48% lower all-cause mortality, 16 to 39% lower cancer mortality and 22 to 57% lower cardiometabolic mortality. Regularity predicted all-cause mortality more strongly than sleep duration [2].

Other studies point in the same direction. In a Swedish study of 43,880 adults followed for 13 years, adults under 65 who consistently slept 5 hours or less had 65% higher mortality than those who consistently slept 6 to 7 hours, but those who slept little on weekdays and longer at weekends did not differ from the reference group [8]. Among 433,268 UK Biobank participants, people describing themselves as definite evening types had 10% higher all-cause mortality than definite morning types [9]. Among 74,862 US nurses followed for 22 years, those who had worked rotating night shifts for 6 years or longer had 11% higher all-cause mortality and 19 to 23% higher cardiovascular mortality than those who never had [10]. The biology of the internal clock is described in sleep physiology.

A study of 172,321 US adults in the National Health Interview Survey scored five low-risk sleep factors: sleeping 7 to 8 hours a night, difficulty falling asleep no more than twice a week, trouble staying asleep no more than twice a week, no sleep medication, and feeling well rested on at least five days a week [11]. Over a median of 4.3 years, 8,681 participants died. Compared with people who had none or one of the factors, those with all five had 30% lower all-cause, 21% lower cardiovascular and 19% lower cancer mortality. Life expectancy at age 30 was estimated at 4.7 years longer for men and 2.4 years longer for women with all five factors, and 7.9% of all deaths were attributed to suboptimal sleep patterns [4]. The quality aspects of this score are described in sleep quality.

Sleep has also been compared directly with other risk factors. In 19,951 US adults followed for a median of 7.6 years, the American Heart Association's eight cardiovascular health measures ("Life's Essential 8") were scored, with 7 to under 9 hours of sleep counted as ideal. The share of deaths statistically attributed to a less-than-high score was 16.5% for physical activity, 14.7% for nicotine exposure, 11.0% for diet, 5.8% for blood pressure, 5.4% for sleep and 4.7% for blood glucose [12]. Sleep in that study was scored by hours alone, without regularity or sleep disorders.

Obstructive sleep apnea is repeated narrowing or closure of the upper airway during sleep, causing pauses in breathing (apneas) or shallow breathing (hypopneas). Its severity is graded by the apnea-hypopnea index, the number of such events per hour of sleep. An estimated 936 million adults aged 30 to 69 worldwide have mild to severe sleep apnea and 425 million have moderate to severe disease [13].

In the Wisconsin Sleep Cohort, 1,522 adults were tested with overnight polysomnography (a laboratory sleep recording) and followed for 18 years. After adjustment for age, sex and body mass index, severe sleep-disordered breathing (30 or more events per hour) was associated with 3.0 times the all-cause mortality of no sleep-disordered breathing. After excluding participants who had used CPAP (continuous positive airway pressure, a mask that keeps the airway open), the figure was 3.8 for all-cause and 5.2 for cardiovascular mortality [3]. A pooled analysis of 12 studies with 25,760 participants found severe sleep apnea associated with 1.92 times the all-cause mortality and 2.15 times the stroke risk of people without it [14].

A Spanish observational study of men followed for a mean of 10.1 years found fatal cardiovascular events in 1.06 per 100 person-years among those with untreated severe apnea, compared with 0.35 among patients treated with CPAP and 0.3 among healthy men [15]. Treatment was not randomly assigned in that study.

The relationship between insomnia and mortality depends on which symptoms are counted. In 1,969 Australians aged 65 or older followed for a median of 9.2 years, the most severe fifth of combined insomnia symptoms had 26% higher mortality than the least severe. The association was driven by daytime symptoms such as difficulty concentrating and difficulty getting going, which carried 66% higher mortality in the most severe fifth, while night-time symptoms alone were not associated with mortality [16].

Sleep medication use has also been associated with mortality. A pooled analysis of 25 studies with 2,350,093 patients found that users of hypnotic (sleep-inducing) or anxiolytic (anti-anxiety) drugs had 43% higher mortality than non-users, with a hazard ratio of 1.60 for benzodiazepines [17]. People prescribed these drugs often have illnesses that raise mortality on their own, which limits interpretation.

In 7,959 participants of the British Whitehall II study followed for 25 years, with 521 diagnosed dementia cases, sleeping 6 hours or less at age 50 was associated with 22% higher and at age 60 with 37% higher dementia risk, compared with 7 hours. Persistent short sleep at ages 50, 60 and 70 was associated with 30% higher risk, independently of sociodemographic, behavioural, cardiometabolic and mental health factors [5]. Because sleep changes can begin years before dementia is diagnosed, measuring sleep in midlife reduces but does not remove the possibility that early disease causes the short sleep. The wider evidence is covered in dementia risk factors and cognitive aging.

Most studies measure sleep once, by questionnaire, and people commonly misjudge their own sleep length. Reverse causation (illness causing a sleep pattern rather than the reverse) is a particular concern for long sleep: depression, heart failure, cancer and chronic inflammation all lengthen time in bed, which may explain why long sleep is associated with higher mortality in almost every study [1, 7]. The authors of the cardiovascular analysis described both short and long sleep as "predictors, or markers" of cardiovascular outcomes [6]. Long daytime naps show a similar pattern, with naps of an hour or longer associated with 27% higher all-cause mortality [18].

Mental health also runs in both directions with sleep: depression and anxiety disrupt sleep and are themselves linked to shorter lifespan, a topic covered in mental health conditions and lifespan. No randomised trial has assigned people to a sleep length and measured mortality, so the number of years gained or lost per hour of sleep has not been established.

Full deep dive: sleep and lifespan.