U-shaped sleep vs mortality, but how is sleep measured

[kpfleger]

I quick Google Image Search for the query sleep mortality shows the widely reported U-shaped curve, with optimum near 7 hours. Has anyone dug into the studies that have examine this to focus on how sleep duration of the subjects was collected? Or has anyone looked into this enough to find which such studies specifically differentiated between hours spent in bed vs. hours spent actually sleeping as determined by some kind of sensor?

 

My expectation and vague memory is that some fraction of these studies use self-reported sleep durations. Even if all the various watches, clips, ballistocardiography, smartphone-on-mattress, etc. sensors are not perfect at measuring sleep stages, I think some of them are decent approximations for total time spent in a sleeping state vs. awake (though they vary, eg my Emfit QS seems much better at realizing how long it takes me to go back to sleep after getting up in the middle of the night than my Garmin FR235).

 

In my case, I seem to only spend an average of about 80% of my time between going to bed and finally waking actually sleeping. So this makes a big difference in terms of where the optimum is for hrs/night. Eg, if I try to hit 7hrs/night real sleep, then I'm at nearly 9hrs/night in bed, but maybe the studies are actually just reporting associations with in-bed time.

 

I haven't had a chance to dig into the studies myself, so I figured I'd ask if anyone else has looked at any of them in enough detail to know, or found any that specifically used a trustable sensor to determine actual sleep time in the subjects.

 

Karl

 

[Matt]

I don't know, but my feeling is that you should just sleep until you feel rested. If you naturally feel rested at 8 hours sleep, but studies say that you should only sleep 7, I think it would be a terrible idea to restrict sleep on purpose. A better idea would be to look at how you can optimize your health and sleeping habits to get the best quality of sleep and then your sleep duration will adjust accordingly.

[mccoy]

Like all other biological parameters, sleep requirement is not a number, rather a random variable, whose average is about 7.5 hours, according to the studies of Kirk Parsley. He explains how these studies have been arranged, people closed in dark chamber starting with the duration of 14 hours and gradually decreasing.

 

7.5 hours average seems reasonable to me. I believe that's net worth of sleeping time, so pre-sleep relaxation time is ruled out. An interestign podcast. Also, there is intra-individual variabililty in sleep requirements. So, some people need more sleep some need less, but we may need more or less in particular periods or in single nights.

 

https://www.ihmc.us/stemtalk/episode011/

 

 

15:50: Polysomnographs reveal that some people wake up 300 times a night, but say they slept fine.

16:13: You don’t need the same amount of sleep every day. Seven and a half hours is the average amount of sleep we aim for to enhance the immune system.

39:17: Sleep adaptation studies show that the average person living in the Western industrialized lifestyle settles down at needing 7.5 hours of sleep.

43:05: A genetic variant allows some people to sleep less and not suffer sleep deprivation as badly as the average person.

 

Edited July 13, 2018 by mccoy

[kpfleger]

Since posting this I have read the excellent book, Why We Sleep by Matthew Walker. Highly recommended. And available as an audiobook. One of the best books on health relationship to any aspect of lifestyle that I've ever read. Better than any single book on nutrition/eating (including CR) that I've ever read, and I've read many more on that area of lifestyle. Similar for exercise.

He specifically says in this book that the sleep recommendations of the National Sleep Foundation and CDC of 7-9 or 7+ are time-in-bed recs and that modern hunter-gatherers actually sleep 6-7.5 hrs/night. That CDC page cites a couple published studies (that I did not dig into).

I now try to spend 8-9 hours in bed in order to get 6.5-7.5 hours actual sleep.

[My sleep efficiency seems to hover around 80% based on my Emfit QS and on my Garmin Forerunner 235 after correlating it with my old Basis Peak (which I trust more for sleep than the Garmin and which agreed with the Emfit) when switching from the Peak to the Fr235.]

[Gordo]

I'm a huge proponent of deep restorative sleep.  I'll have to check out that book, I like the reviews.

[Michael R]

My recollection is that most of the epidemiological studies that I've seen rely on self-report. PMID: 29415200 pools data from 3 such cohort studies:

Quote

In the FPS, sleep duration was measured by asking participants: ”How many hours do you usually sleep per 24 hours?”, in the WH II “How many hours of sleep do you have on an average week night?” and in the SLOSH study, participants were asked regarding working/weekdays “At what time do you normally go to bed (turn the lights out)?” and “At what time do you normally get up?”.

If you think about the number of years of followup and the number of participants required to get a mortality outcome in  an epidemiological study in the general population, it's hard to imagine doing PSGs on that many people (let alone with multiple data points, since a single night is not likely representative). And people sleep abnormally when in a sleep lab or anywhere other than their normal sleeping environment, so you'd actually get garbage in anyway.

And again going back to the number of years required to get a mortality outcome in  an epidemiological study in the general population, and the additional time required to organize and analyze such a study, there can't really've been any such studies done with fitness trackers to get more real-world sleeping data either: yes there were actimeters, but many of them were unsuited for sleep, and even today most consumer health trackers' sleep algorithms are utter garbage (the dearly-departed Basis Peak being a striking exception).

Speaking of sleep trackers: see this post that I was just finally pushed into making by this thread.

I recommend listening to the Kirk Parsley IHMC podcast linked by McCoy from 39:15-44:12, which describes sleep adaptation studies coming to 7-7.5 h as required for optimal performance: I think it's safe to assume that the deleterious short-term effects on the body would lead to accumulating damage.

Edited October 13, 2018 by Michael R

[Michael R]

This is from the "Discussion" from a recent Meta‐Analysis of Self‐Reported Sleep Duration and Quality and Cardiovascular Disease and Mortality (PMID 30371228):

Quote

Several studies have evaluated the mismatch between self‐reported sleep and more‐objective measures. An analysis of 669 participants in the CARDIA (Coronary Artery Risk Development in Young Adults) Sleep Study suggests a moderate correlation between self‐reported and objectively measured sleep duration with evidence of systematic errors in the mean and calibration.⁹⁷ Another study of 2086 Hispanic Americans suggests a moderate correlation between self‐reported sleep and actigraphy results.⁹⁸ A smaller study that compared self‐reported sleep and actigraphy in 56 participants reported a poor agreement between sleep duration and quality, as assessed by a questionnaire and objectively measured sleep.⁹⁹ In addition, a different analysis of 63 patients who were overweight and obese individuals and who underwent actigraphy also reported a weak correlation between usual sleep time and actigraph estimates.¹⁰⁰ With respect to CVR, studies suggest that self‐reported sleep duration is different from objectively assess sleep duration.¹⁰¹ Although the exact reasons for discrepancies between subjective and objective measures are not clear, it appears that self‐reported sleep is only, at best, moderately correlated with objective measures of sleep and perhaps multiple measures of sleep duration should be recommended for future studies.

The meta-analysis itself finds:

Quote

Seventy-four studies including 3 340 684 participants with 242 240 deaths among 2 564 029 participants who reported death events were reviewed. ... Self-reported duration of sleep >8 hours was associated with a moderate increased risk of all-cause mortality, with risk ratio , 1.14 (1.05-1.25) for 9 hours, risk ratio, 1.30 (1.19-1.42) for 10 hours, and risk ratio, 1.47 (1.33-1.64) for 11 hours. No significant difference was identified for periods of self-reported sleep <7 hours, whereas similar patterns were observed for stroke and cardiovascular disease mortality. Subjective poor sleep quality was associated with coronary heart disease (risk ratio , 1.44; 95% confidence interval, 1.09-1.90), but no difference in mortality and other outcomes.

Conclusions Divergence from the recommended 7 to 8 hours of sleep is associated with a higher risk of mortality and cardiovascular events. Longer duration of sleep may be more associated with adverse outcomes compared with shorter sleep durations.

... but long self-reported sleep duration may reflect COPD, sleep disturbed by stress, noise, and light, etc.

[Genny]

Cause-effect is backwards in these studies except for a very few people.  You can't make yourself live longer by forcing yourself to sleep more or less--that doesn't make sense--but you probably can get some effect if you're a sleep-deprived zombie and sleep a bit more, or if you address the health issues that are making you sleep too much or too little.

[Michael R]

 

On 2/2/2019 at 1:13 PM, Genny said:

Cause-effect is backwards in these studies except for a very few people.  You can't make yourself live longer by forcing yourself to sleep more or less--that doesn't make sense--but you probably can get some effect if you're a sleep-deprived zombie and sleep a bit more, or if you address the health issues that are making you sleep too much or too little.

I've been meaning to get back to this for some time.

I would agree to a point that "You can't make yourself live longer by forcing yourself to sleep more," if by "force" you mean an individual who (hypothetically) is  perfectly well-rested with just 6 h of sleep opportunity time (actually a vanishingly small number of people) somehow willing hirself to sleep an extra hour just to make an epidemiologically-determined cutpoint.

But as is I think widely recognized, most people have adopted habits that restrict their sleep to less than what is actually needed to meet their physiological needs staying up late (relative to a set wake time) to complete work or study projects, or to go out for entertainment, or to watch TV, or to keep up with the Kardashians on Instagram, or to address the unending problem that someone on the internet is wrong. Very few people are actually getting recommended amounts of sleep (meaning, actually, sleep-opportunity time, not PSG-measured sleep), as study after study documents, and multiple studies show that when taken away from other stimuli and from clocks — or even just blue light from devices — they sleep more and better than they habitually allow themselves to.

Studies also show that people are very poor judges of whether they're getting enough sleep, or how sleep-deprived they are. In PMID 12683469, "48 healthy adults (ages 21-38)" were randomized to " 8 h, 6 h or 4 h [of sleep] per day for 14 days, or to 0 h for 3 days". "Each study also involved 3 baseline (pre-deprivation) days and 3 recovery days." "Chronic restriction of sleep periods to 4 h or 6 h per night over 14 consecutive days resulted in significant cumulative, dose-dependent deficits in cognitive performance on all tasks" — that is, the longer you suffered these very mild sleep deprivations, and the shorter your sleep, the worse your cognitive performance got:

Despite the fact that these deficits (PVT, DSST, SAST) got worse and worse over time and with lower and lower amounts of sleep, "Sleepiness ratings suggest that subjects were largely unaware of these increasing cognitive deficits": the subjects' self-reported sleepiness ratings flattened after the first 3-4 days (SSS in the Figure above), "and did not significantly differentiate the 6 h and 4 h conditions." "[C]hronic restriction of sleep to 6 h or less per night produced cognitive performance deficits equivalent to up to 2 nights of total sleep deprivation, [so] it appears that even relatively moderate sleep restriction can seriously impair waking neurobehavioral functions in healthy adults." The disconnect between self-reported sleepiness and objective deficits "may explain why the impact of chronic sleep restriction on waking cognitive functions is often assumed to be benign."

Here are a couple of trials of "forcing" people to sleep more, by simply making them go to bed earlier than they normally do and not waking them up until they had achieved the target sleep time:

In PMID 21731144,

Quote

Eleven healthy students on the Stanford University men's varsity basketball team (mean age 19.4 ± 1.4 years) ... maintained their habitual sleep-wake schedule for a 2–4 week baseline followed by a 5–7 week sleep extension period. ... with a minimum goal of 10 h in bed each night. ...  Total objective nightly sleep time increased during sleep extension compared to baseline by 110.9 ± 79.7 min [that is, they "forced" themselves to get almost two hours more actual sleep than they did during a 2-4 week baseline period] (P < 0.001). Subjects demonstrated a faster timed sprint following sleep extension (16.2 ± 0.61 sec at baseline vs. 15.5 ± 0.54 sec at end of sleep extension, P < 0.001). Shooting accuracy improved, with free throw percentage increasing by 9% and 3-point field goal percentage increasing by 9.2% (P < 0.001). Mean [Psychomotor Vigilance Task] reaction time and Epworth Sleepiness Scale scores decreased following sleep extension (P < 0.01). [Profile of Mood States] improved with increased vigor and decreased fatigue subscales (P < 0.001). Subjects also reported improved overall ratings of physical and mental well-being during practices and games.

In PMID: 31166059, the authors identified "Seven studies that aimed to increase sleep duration in adults by any sleep extension intervention and described at least one cardiometabolic risk factor ... These studies had a combined sample size of 138 participants who were either healthy (n = 14), healthy short‐sleeping (n = 92), overweight short‐sleeping (n = 10), or pre‐ or hypertensive short‐sleeping (n = 22) individuals. The durations of the sleep extension interventions ranged from 3 days to 6 weeks, and all successfully increased total sleep time by between 21 and 177 min. Sleep extension was associated with improved direct and indirect measures of insulin sensitivity, decreased leptin and peptide tyrosine‐tyrosine, and reductions in overall appetite, desire for sweet and salty foods, intake of daily free sugar, and percentage of daily caloric intake from protein."

 

It appears that there are benefits to sleep extension in children, too.

So, yes: most people can "force" themselves to get more sleep, by forcing themselves to go to bed early enough to get adequate sleep — a hard thing, with all the opportunities or entertainment and procrastination presented by our modern world. The act of going to bed is more likely to lead to more and better sleep when enabled by good sleep hygiene (compare the CDC , the American Academy of Sleep Medicine, and the National Sleep Foundation) and some more high-tech tools, like good blue-blocking glasses — and for those that need it, cognitive-behavioral therapy for insomnia (CBT-I).

And the evidence is that even in the very short term, "forcing" yourself to get more sleep is good for your performance and health.

[Ron Put]

As stated by others, most people overestimate their sleep duration (as they underestimate their caloric intake).

There have been some studies aggregating measured sleep duration and mortality rates, but those are still rare. Here is one which appears rather well-supported:

"Background

The folk belief that we should sleep 8 hours seems to be incorrect. Numerous studies have shown that self-reported sleep longer than 7.5 hours or shorter than 6.5 hours predicts increased mortality risk. This study examined if prospectively-determined objective sleep duration, as estimated by wrist actigraphy, was associated with mortality risks.

Methods

From 1995–1999, women averaging 67.6 years of age provided one-week actigraphic recordings. Survival could be estimated from follow-up continuing until 2009 for 444, with an average of 10.5 years before censoring. Multivariate age-stratified Cox regression models were controlled for history of hypertension, diabetes, myocardial infarction, cancer, and major depression.

Results

Adjusted survival functions estimated 61% survival (54%–69%, 95% C.I.) for those with sleep less than 300 min and 78% survival (73%–85%, 95% C.I.) for those with actigraphic sleep longer than 390 min, as compared with survival of 90% (85%–94%, 95% C.I.) for those with sleep of 300–390 min. Time-in-bed, sleep efficiency and the timing of melatonin metabolite excretion were also significant mortality risk factors."



https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3010336/#!po=1.21951

[TomBAvoider]

This study examined if prospectively-determined objective sleep duration, as estimated by wrist actigraphy, was associated with mortality risks.

I don't find this compelling. "Objective sleep duration as estimated by wrist actigraphy". When I click on the appendix re: wrist actigraphy, I don't find nearly enough detail to assess the validity of "wrist actigraphy" as estimating sleep duration. What I did find, was this sentence:

"As compared to polysomnography, actigraphic methods may over-estimate or under-estimate the amount of sleep, depending on the actigraphic methods and characteristics of the subject sample."

There is no detail supplied as to which particular actigraphic methods were used in this study. The fact that there are many, and that they are subject to dispute with over and under estimating sleep duration when compared to other methodologies, does not inspire any confindence in me at all.

In fact, I find it pretty meaningless to speak of "sleep duration" when you cannot give me an exact definition of what the sleeping state is. One study may use one method another may use a different method. To then draw conclusions about "duration" of such a fuzzy concept strikes me as suspect.

My personal experience of sleep is such that I would find it very difficult to estimate my own sleep duration. One of the biggest reasons is that I cannot with certainty establish when it is that I have entered sleep or am in a sleep state. Sometimes it seems to me that I'm awake in bed, but when I examine my thought process I realize it is nonsensical and indicates that I'm actually in a sleep-like state. To complicate matters even more, my sleep has changed with age. I don't sleep the same way - it seems my sleep is in some ways shallower than when I was younger, or more fragmented etc. I'm sure one could measure something like brain waves and other biomarkers, but then we run up against definitional issues yet again. What is sleep? And how can you measure sleep duration, when the character of sleep has changed? Is an hour of "sleep" at age 20 the same as an hour of "sleep" at age 60? I strongly feel that it's not equivalent. If so, then what point is there in saying X number of hours of sleep is optimal - what sleep, with what characteristics? Five hours of sleep by a 20 year old may not be at all the same in effect as five hours of sleep by a 60 year old - of that I'm pretty sure. 

Bottom line, I feel there is too much we don't know and have not fully characterized to be able to sensibly measure. Therefore a general number as bandied about in these studies "X number of hours/minutes of sleep duration is optimal wrt. all-cause mortality" - is almost meaningless.

Human physiology changes with age. Surely we don't think that there is an "optimal" number of calories, or macronutrient composition for a diet to be optimal for all - we recognize that a child needs a different diet than an adult, and in fact there are many studies that recommend different diet compositions for the old and very old. Why is sleep any different? I think it isn't and merely speaks to our ignorance. Once we learn more, I'm sure it will all be much more nuanced (not to mention individual differences may emerge too - just as with diet). Making blanket statements about "optimal duration" that's supposed to fit all, strikes me as unjustified.

Now I'm not saying that there aren't various hypesters who want to set themselves up as gurus and sell books on a trendy topic of sleep (see: Matthew Walker), but as we've seen with diet gurus, there's no sense in jumping on someone's bandwagon, because the science isn't there (as yet). Show me the science! What I see so far, including in this thread, I find highly unpersuasive. YMMV.

[Ron Put]

On 6/17/2019 at 7:34 PM, TomBAvoider said:

This study examined if prospectively-determined objective sleep duration, as estimated by wrist actigraphy, was associated with mortality risks.

I don't find this compelling. "Objective sleep duration as estimated by wrist actigraphy". When I click on the appendix re: wrist actigraphy, I don't find nearly enough detail to assess the validity of "wrist actigraphy" as estimating sleep duration. What I did find, was this sentence:

"As compared to polysomnography, actigraphic methods may over-estimate or under-estimate the amount of sleep, depending on the actigraphic methods and characteristics of the subject sample."

 There is no detail supplied as to which particular actigraphic methods were used in this study. The fact that there are many, and that they are subject to dispute with over and under estimating sleep duration when compared to other methodologies, does not inspire any confindence in me at all....

Well, you may not find it compelling, but it's one of the best studies around I am aware of. Of course they will have a disclaimer about the accuracy of wrist-based actigraphy compared to a lab sleep study setting, which measures your your brain waves and eye movement, among other things. Actigraphy may not be as accurate, but it's still pretty good and a great deal more accurate than self-reporting, which most other studies use.

It also jives with another, more recent study in protected tribal subjects, also using actigraphy, which found that the average amount of sleep they get was about 6 hours and 20 minutes or so, and that their waking time was not light-dependent, but it corresponded to the coldest time in the morning. (I don't have the time to find it up now.)

[kpfleger]

Ron, thank you for the https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3010336/#!po=1.21951 study. This is the best direct answer yet to the question I posed in the 1st post. As Michael said, it will be some time before we have lots of studies using actual sleep measurement and long-term health followup because we didn't have measurement a long time ago and what measurement we had (sleep studies) were very sleep disruptive. We obviously didn't have FitBits and Apple Watches in the 1970s, but even if whatever wrist actigraphs we did have were systematically different than what people use now, it is clearly a strict improvement on self-reported time-in-bed, and the study demonstrates exactly what I hypothesized in the 1st post:

Quote

This study confirms a U-shaped relationship between survival and actigraphically measured sleep durations, with the optimal objective sleep duration being shorter than the self-report optimums.

It's notable that getting too little sleep is much worse than getting too much, based on the pasted graph. It seems odd that the highest breakpoint they used was 6.5hrs. Given the other science suggesting 7-7.5hr as optimal (the studies Kirk Parsley described in the linked podcast above) or 6-7.5hr as optimal (some of the modern hunter-gatherer studies described in Matthew Walker's book I recommended above), it seems strange they wouldn't separate the above 6.5hr group with a breakpoint at ~7.5hr. I haven't had a chance to read the paper yet, but will look at this when I do. Maybe there weren't enough subjects in the above 7.5hr range. But if there were, one hopes their outcomes don't drag down those in the 6.5-7.5hr range.

Tangentially, I note that Dreem's website homepage says the Dreem2 is available now, but when you click through to the order page, it still says not available yet and has a place to register your email to be notified when that changes.

Karl

 

[kpfleger]

This is surprising: the study did report results at finer-grained breakpoints (8 sleep ranges instead of 3), but it was not a U-shaped curve with mortality. The lowest and highest sleeping groups survived much better than the 2nd lowest and 2nd highest, and not that much worse than the groups in the middle ranges (with confidence intervals that don't go way into bad territory so it doesn't look like just randomness due to low n). The differences between adjacent groups may not have been statistically significant, but when graphed the overall 8-group bar-chart still represents a striking departure from what you would expect for a U-shaped dose-response (see attached bar chart, fig 1 of the paper).

In numbers, those sleeping >7.5hr (n=15) had 89% survival (95% CI 81-91%) vs those sleeping 7-7.5hr (n=31) having 58% (45-71%). The 81-91% range for >7.5hr doesn't seem that much worse than the 85-94% CI for the pooled 5-6.5hr group. If this were some weird quirk due to the low n, I would expect a bigger CI that ranged down further into bad survival %s.

I didn't see any discussion of this turning down of the ends of the U curve in glancing very briefly through the later parts of the paper. But this doesn't match any of the other sleep research I've ever heard about. It seems so odd, it's hard for me to take this paper as a reason to think that >6.5hr is the point at which more starts to become bad, as suggested by their pooled 3-group analysis (and the figure 2 that Ron posted above). So until I see some study that replicated this, I'm still going to go with 7.5hr of actual sleep measured by a tracking device as the rough amount where more sleep may start to become a negative, as suggested by the hunter gatherer research reported in the Walker book and the studies described by Parsley.

My own sleep averages around 6.5hrs (measured w/ Emfit QS under mattress + Garmin Fr235 at wrist).

Karl

[mccoy]

There should be a rational mechanistic explanation for the anomaly depicted in the graph. In lieu of reasonable mechanistic explanation (and I have difficulties finding one), IMO it's just an anomaly, due to statistical variability of subgroups, measurement errors, confounding factors and so on.