Lifespan linked to sleep

Wednesday May 5 2010

“Not enough sleep leads to a wake,” the Mirror reported. It said that sleeping less than six hours a night makes you 12% more likely to die prematurely than someone who sleeps up to eight hours.

This is a large, well-conducted review of studies looking at the relationship between sleep duration and death. Its findings do not mean that not getting the standard amount of sleep will result in an early death. Most of these participants were over 60 when they began, and the different studies varied in length between four and 25 years.

Though the reviewers did find that six or less hours of sleep was associated with a 12% increased risk of death, they also found a 30% increase linked with nine or more hours. It is unclear why the newspapers all focused on the risks of less sleep.

A causal relationship is possible, and more plausibly so in cases of extreme sleep deprivation. However, both sleep and lifespan are significantly affected by physical and mental health, and sleep duration might be a better general indicator of health, rather than risk of death.

Where did the story come from?

The research was carried out by Dr Francesco Cappuccio and colleagues from the University of Warwick and University of Naples Medical School. The study was funded in part by an EC Grant. The study was published in the peer-reviewed scientific journal Sleep .

The newspapers have all focused on the finding that six or less hours of sleep a night is associated with an increased risk of early death. However, the research found that people sleeping fewer hours were no more at risk than people sleeping nine or more hours a night. In fact, people sleeping for longer had a 30% increased risk compared to people sleeping less who had only a 12% increased risk.

What kind of research was this?

This was a systematic review and meta-analysis. The researchers systematically searched for and combined the results of cohort studies that investigated whether there is an association between sleep duration and death from any cause. This type of review is the best way of assessing the current evidence for a potential association – in this case, between duration of sleep and death. Reviews like this do have some limitations however, in that the overall calculated risk from the combined studies is likely to be affected by differences in their methods, follow-up and assessment of outcomes.

What did the research involve?

The authors searched several major medical databases for all published cohort studies on sleep duration and death from any cause (all-cause mortality). To be eligible, studies had to have assessed at their start how much sleep participants were getting. Death had to be the main outcome of interest, and the studies had to have followed participants for at least three years, and to have categorised sleep into different set durations. Independent investigators assessed the quality of the studies and selected them for inclusion.

Sleep categories varied between studies:

  • Standard sleep (the reference category): seven hours (most studies), 6-8 hours, 7-8 hours, 7-9 hours, nine hours.
  • ‘Short’ sleep: four hours or less, five hours or less, six hours or less.
  • ‘Long’ sleep: nine hours or more, 10 hours or more, 12 hours or more.

The risk ratios for death by sleep category were obtained from the individual studies, and a combined risk was calculated. Statistical methods were used to take into account the differences between the studies. The researchers also carried out sub-group analyses by age, gender, socioeconomic status, definitions of sleep duration, and length of follow-up.

What were the basic results?

The researchers identified 16 suitable studies, in 27 cohorts (groups). These covered a total of 1,382,999 adults from eight different countries. Two of the studies investigated women only, and the other 14 were mixed gender. Duration of follow-up varied between four and 25 years. All studies assessed sleep duration by questionnaire, and all assessed the outcome of death through death certificates. Across the studies, there were 112,566 deaths during follow-up.

Of the 16 studies, 15 (involving 25 cohorts; 1,381,324 adults; and 112,163 deaths) were suitable for inclusion in a meta-analysis looking at the association between short sleep and death. It found that short sleep (six hours or less) was associated with a 12% increased risk of death during follow-up (relative risk 1.12; 95% CI 1.06 to 1.18).

All 16 studies were suitable for inclusion in a meta-analysis, which looked at the association between long sleep and death. Long sleep (nine or more hours) was associated with a 30% increased risk of death during follow-up (relative risk 1.30; 95% CI 1.22 to 1.38).

Sub-analysis revealed that the association between short sleep and death was not affected by sex, age or socioeconomic status. But with long sleep, the risk association was stronger in older people, and with increasing duration of sleep up to more than 10 hours.

How did the researchers interpret the results?

The researchers conclude that both short and long sleep durations are “significant predictors of death in prospective population studies”.


This well-conducted systematic review examined the association between sleep duration and all-cause mortality. It has numerous strengths, one of which is that it explicitly outlined its review methods. It also included a large number of studies in a total of 1,382,999 adults, capturing 112,566 deaths during follow-up. However, there are some unavoidable limitations due to the nature of the research question and the studies included.

  • Meta-analyses combine similar studies, but these have often used different methods. In this case, the definitions of what was considered to be short, standard, and long sleep durations varied between studies. Standard sleep duration was typically seven hours, but ranged from six hours to nine hours. Definitions of short sleep varied from less than four hours to six hours. Similarly, follow-up varied from four to 25 years. The possible confounders that the studies were adjusted for also varied.

On the positive side, however, all of the studies used the same methods for examining sleep (by questionnaire)  and death (by death certificate). In addition, the researchers have explicitly stated the statistical methods they used to take into account the differences between the studies.  However, in taking this into account they calculated that there was indeed significant difference between the studies (known as statistical heterogeneity), as may have been expected, given that there was such variation between studies in their sleep categories and length of follow-up. * The assessments of sleep duration were by self-report (questionnaire), and were only measured once at the beginning of the studies. This one time measurement probably does not represent a person’s life-long sleeping pattern. Also, as the researchers say, the respondents’ answers may have differed in that they may have considered sleep to mean all time in bed, time actually asleep in bed, or all time asleep throughout the day (including naps). * The individual studies do not appear to have taken into account the participants’ mental or psychological health, for example, whether they had depression or stress. These are known factors for affecting both sleep duration and overall mortality. * Most of the studies were in people who were over 60 years of age at enrollment, and they looked at the risk of death during follow-up, ranging from four to 25 years. Therefore, newspaper headlines that shorter sleep duration is linked to ‘early death’ may need to be revised, as no age bracket can be easily attached to this. Very different results may have been obtained if the studies had assessed younger or middle-aged populations. * Although the newspapers focused primarily on the risks of ‘short’ sleep, and the 12% increase in risk of death, ‘long’ sleep was associated with a 30% significantly increased risk of death. This implies that there is an increased risk of death from anything other than seven or eight hours of sleep a night.

This is interesting and informative research. However, it should not be taken to mean that people who do not follow the ‘standard’ pattern for sleep are more likely to die early. Though a causal relationship is possible, the underlying reasons for poor sleep patterns and their possible relation to physiological changes in the body also need consideration. The BBC quotes Professor Horne from the Loughborough Sleep Research Centre: "Sleep is just a litmus paper to physical and mental health. Sleep is affected by many diseases and conditions, including depression."

In addition, different people need different amounts of sleep, and this can be influenced by age, lifestyle, diet and environment. For example, newborn babies can sleep for 16 hours a day, while school-age children need an average of 10 hours sleep. Most healthy adults sleep for an average of seven to nine hours a night. As you get older, it is normal to need less sleep. Most people over 70 need less than six hours sleep a night, and they tend to be light sleepers.

Analysis by Bazian
Edited by NHS Choices