select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to observational study on association of sleep duration and sleeping pill use and with mortality and life expectancy

An observational study published in Sleep Health looks at the association between sleep duration and sleeping pill use, with mortality and life expectancy.


Dr Baptiste Leurent, Lecturer in Medical Statistics, University College London, said:

“This is a well conducted analysis, comparing the all causes-mortality for different sleeping pattern groups, and users of sleeping pills. Unfortunately, the press release seems to overstress the significance of some of the findings from the paper.

“The unadjusted results are indeed quite striking, with 8.9% of participants taking sleeping pill at the start of the study who died during the follow-up period (of around 10 years), compared to 3.4% in the non-pills takers. But this is of course partially explained by confounders (e.g. those taking sleeping pills were older).

“The researchers conducted adjusted analysis (adjusting for age, sex, medical history etc.), and still find an increase in mortality in sleeping-pills taker, but it was not necessarily statistically significant. It is difficult to give an overall number, as they did not conduct a global analysis comparing all the “sleeping pills taker” to the “non-takers”.  But in the optimal 6-8 hours sleeper group, they found a non-significant 10% increase in mortality (hazard ratio =1.10, Confidence interval 0.91 to 1.32) in pill takers after full adjustment.  In is not clear why the results reported in the press release focus on a partially adjusted model (not adjusted for psychiatric medication), for example when reporting a “55% higher mortality” or “may increase mortality risk by more than half”.

“It is also very unclear where “may more than double your mortality risk” came from, and this is probably an over-statement. Looking at the fully adjusted models (or even the partially adjusted), none of the four sleeping groups had double the hazard, for pill takers vs non-pill takers. This doubled in mortality was perhaps from comparing those taking pill and sleeping over 8 hours, to those non taking pill and sleeping 6-8 hours. But this comparison is of limited relevance.

“The study was well conducted statistically, but included several limitations. It is worth noting that the sleep data were collected over 20 years ago, and the mortality follow-up completed over 10 years ago. Also, around 20% of patients from the original cohort did not have sufficient sleep data to be included. It is not clear why this was the case, and if these excluded participants could differ from those included (e.g. having less sleep problems?). Finally, as with any observational study one should be very cautious about making causal statement. It is likely that some further confounders could explain the apparent association. For example, the analysis was not adjusted for “having difficulty falling asleep”, which looks like an obvious confounder in this association.”


Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“The press release for this study does broadly match what is in the research paper, although the quote from the lead author does need careful interpretation. And the press release doesn’t mention any of the several important limitations of the research, which are clearly spelled out in the research paper. 

“The first limitation, and I’d say the most important, is that this is an observational study, and so there are difficulties in knowing what causes what. The researchers found that people who took sleeping pills had a higher risk of early death than people who did not take them. That could be because the sleeping pills caused the increased risk, but that’s far from the only possible explanation for the findings. Some other factor might, independently, cause people to be more likely to take sleeping pills, and also increase the risk of early death. In that case stopping taking the sleeping pills, or never starting on them, might not improve a person’s life expectancy at all. 

“The researchers did make statistical adjustments and extra analyses to allow for some factors that might be involved in this way. For instance, they adjusted for several health conditions that might have independently affected the chance of being a sleeping pill user and the chance of an early death. These adjustments do go some way to increasing confidence that it’s the sleeping pills that are responsible for the differences in mortality. But, as the researchers say in their paper, doubt about cause and effect still remains. I’m certainly not saying that taking sleeping pills has no effect on mortality – only that this research can’t rule out that possibility. 

“That’s why one needs care in interpreting the lead author’s statement quoted in the press release, that “taking them [sleeping pills] may more than double your mortality risk.” You might read this as meaning that the increase in risk might not quite be more than double, but in fact it includes the possibility that taking the pills could have no effect on mortality at all. 

“The research paper lists several other limitations too. For instance, the measures of sleep were entirely based on what the participants reported to the researchers, and the researchers point out that such subjective measures of sleep are not all that strongly correlated with objective measures. And duration of sleep and use of sleeping pills were recorded only at the start of the study for most participants. These things are likely to change over time, for instance over the 10 years on average for which people were followed up, but the study couldn’t take that into account. Indeed the life expectancy calculations in the research involve the assumption that people’s sleep duration and use of sleeping pills wouldn’t change right from when they entered the study, at average age 40, until they died possibly many decades later. So it’s hard to tell how much reliance to put on those specific life expectancy figures. 

“Another point I’d make us that all the mortality calculations have to rely on those who died during the 10 average years of follow-up, and in fact only 3.5% of the participants did die during the study. 

“Finally, another point of caution is that social determinants of sleeping pill use are likely to vary a lot between different societies, so it’s possible that the results of a similar study in the UK could be very different from the results of this study on people in Taiwan. Indeed the researchers point out that their participants aren’t even very typical of the general population of Taiwan, because they were people who had chosen to pay a fee for health screening.”



‘Association of sleep duration and sleeping pill use with mortality and life expectancy: A cohort study of 484,916 adults’ by Yu Sun et al. is published in Sleep Health: Journal of the National Sleep Foundation (an Elsevier journal)




Declared interests

Dr Baptiste Leurent: “No conflict of interest.”

Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee.  My quote above is in my capacity as an independent professional statistician.”

in this section

filter RoundUps by year

search by tag