Sleep sits in an unusual relationship to disease. It can be a cause of illness, a consequence of illness, and a symptom of it\u2014sometimes all three at once.1 Losing sleep can worsen mental health, just as depression can disrupt normal sleep. Chronic pain keeps you awake, and sleeplessness worsens pain. Sleep apnea doesn\u2019t just coincide with sleep loss\u2014sleep disturbance is an inescapable feature of the disorder. Heart failure, cancer, and dozens of other conditions perturb sleep long before and long after any diagnosis is reached.\u00a0<\/p>\n
Which brings us to a core limitation of observational research: it can show that two things travel together, but not which one is steering. Epidemiologists call this trap reverse causation\u2014when the outcome you\u2019re studying (illness, and eventually death) is actually driving the exposure you\u2019re measuring (sleep disruption), rather than the other way around. And the list of things that independently disturb sleep and shorten life is effectively endless, which means no amount of statistical adjustment ever fully clears it away. So when we observe short sleepers dying at higher rates, we are looking at a knot this kind of data cannot untie.<\/p>\n
A recent meta-analysis has caught the attention of health and longevity spaces, reporting that across 79 observational studies, sleeping fewer than 7 hours per night is associated with a 14% increase in all-cause mortality.2 The authors, and many commentators, extrapolate beyond what is possible from these sorts of observations: they claim that a lack of sleep is causing an increase in mortality. Given what we know about sleep loss, that conclusion makes sense\u2014but being reasonable isn\u2019t the same thing as being proven.<\/p>\n
The most telling detail is one the authors underplay. Long sleepers\u2014people clocking nine or more hours per night\u2014showed an even larger jump in mortality: 34%. The issue here is that there is no compelling biological mechanism by which simply sleeping an extra hour would substantially shorten life\u2014the strongest causal evidence base for negative outcomes comes from studies of sleep loss, not sleep surplus.\u00a0\u00a0<\/p>\n
When the bigger effect size is the one with no mechanism behind it, that\u2019s a strong hint the entire signal is tangled up with underlying illness. The far simpler explanation is that long sleep is a symptom: the sick, the frail, and the not-yet-diagnosed tend to sleep more.\u00a0<\/p>\n
This study cannot tell us how much of that 14% increase in mortality is caused by a lack of sleep and not by diseases that cause people to lose sleep. It is therefore not very useful for estimating the causal effect of sleep duration on mortality. So how could we get a better look at causality here\u2014how can we know when sleep loss is the driver that brings mortality along for the ride, rather than the other way around?<\/p>\n
Even the gold standard isn\u2019t good enough<\/p>\n
There is a method designed precisely to cut through reverse causation in epidemiological data, which many of you are familiar with by now: Mendelian randomization, or MR. The idea is elegant: your genes are fixed at conception and shuffled essentially at random, long before any disease appears. So if we can identify gene variants that nudge people toward shorter sleep, we can use those variants as a kind of natural randomized experiment. We can \u201csort\u201d people into short sleep and normal sleep groups by genes, let nature run its experiment, and see if we find an association in mortality across an entire lifespan. We can get a clue about causality from observational data by using genetic variation.\u00a0<\/p>\n
When researchers run this analysis, genetically predicted short sleep does predict cardiovascular disease\u2014higher rates of hypertension, heart attack, and coronary disease. And the long-sleep association mostly evaporates under MR, losing its statistical significance, exactly as the \u201csymptom, not cause\u201d reading predicts. So far, so good for taking short sleep seriously.<\/p>\n
But with sleep, there\u2019s a catch. MR only works if the \u201csleep genes\u201d affect disease only through sleep. If a gene influences both sleep and the heart by separate routes, the method quietly breaks\u2014a problem called horizontal pleiotropy (one gene, several effects).\u00a0<\/p>\n
For sleep, that isn\u2019t a hypothetical worry: it\u2019s the expected case. Biology likes to use the same genes and pathways for numerous functions, especially when it comes to the brain. And the biology that sets how long we sleep runs through metabolism, arousal, and stress hormone pathways\u2014the very same systems that govern the heart. The reasonable default assumption is that anything tugging on sleep also tugs on each of these systems directly. And each of these systems affects how long we live.<\/p>\n
Consider a real example. A variant in a gene called ADRB1 shortens how long its carriers sleep by making certain wake-promoting neurons in the brainstem easier to switch on.3 ADRB1 encodes the \u03b21-adrenergic receptor\u2014critical for wakefulness. But ADRB1 is also the main adrenaline sensor on heart muscle: it is the target of the beta-blockers cardiologists prescribe every day. So a genetic link between \u201cshort sleep\u201d and heart disease running through ADRB1 could just as easily be the receptor acting on the heart directly. Sleep disruption may simply be accompanying a stressed cardiovascular system rather than causing it.\u00a0<\/p>\n
In principle, it is possible to rule this out. One would do so through a multi-year, dedicated research program on a single gene. A recent MR study on sleep uses 78 different genes to find its effect, most of which we know next to nothing about.4 To rule out horizontal pleiotropy\u2014to rule out the possibility that these genes are working on sleep and something else at the same time\u2014each gene would need to be individually dissected to prove its path to mortality runs through sleep alone and nothing else. Across dozens of variants, no one has done this and it\u2019s unlikely anyone ever will.\u00a0<\/p>\n
So MR solves the reverse-causation problem, that is true. But for sleep studies, it immediately hands us a new one: horizontal pleiotropy. Which means even our best tool can\u2019t realistically tell us whether short sleep itself shortens life.\u00a0<\/p>\n
So what can we do when even our gold standard isn\u2019t good enough?<\/p>\n
When perfection isn\u2019t possible<\/p>\n
The study that would actually settle this\u2014randomly assigning thousands of people to a lifetime of short sleep or adequate sleep then counting the deaths\u2014is neither ethical nor feasible.\u00a0<\/p>\n
So we change the question\u2014the good experiment you can do is better than the perfect one you can\u2019t. Instead of asking how a lifetime of short sleep affects overall mortality, which seems effectively unanswerable, we ask something we can answer: How does short-term sleep loss measurably affect the systems we care about for longevity?\u00a0<\/p>\n
How does sleep loss affect the systems we care about?<\/p>\n
Here the evidence is genuinely causal, because these studies are true experiments. Restrict someone\u2019s sleep for a few days, watch the biomarkers move, let them recover, and watch how the markers move back. No genetics required.<\/p>\n
The short answer: insufficient sleep tips the body into a pro-stress, pro-catabolic state\u2014one that breaks tissue down rather than building it up. And the effects arrive fast.<\/p>\n
After even a single night of deprivation, cortisol\u2014the body\u2019s chief stress hormone\u2014climbs, while testosterone and growth hormone fall. In one controlled experiment, a single night of total sleep loss raised cortisol by 21%, dropped testosterone by 24%, and cut the rate of muscle protein synthesis by 18%. This single night produced measurable anabolic resistance: the same protein meal now built less muscle.5 In other words, the same exercise and protein intake produced less adaptation. The consequences for body composition are concrete. In dieters, restricting sleep to 5.5 hours a night for two weeks caused roughly 60% more of the weight they lost to come from lean muscle rather than fat\u2014on identical calories.6 Sleep, in other words, helps decide what you lose when you lose weight.<\/p>\n
The metabolic and cardiovascular systems shift too. A few days of restricted sleep impairs glucose tolerance and insulin sensitivity in healthy young people, while blood pressure, sympathetic (\u201cfight-or-flight\u201d) nervous activity, and inflammatory markers all rise.7\u00a0<\/p>\n
These aren\u2019t obscure lab readouts\u2014they are the same intermediates we already track when assessing someone\u2019s risk for metabolic and heart disease. Sleep is a lever on body composition, metabolism, and mood, and deprivation pulls in the wrong direction across nearly every axis at once.<\/p>\n
Tracking what matters<\/p>\n
Population-level studies will keep struggling to prove that short sleep shortens life, because they are structurally unable to separate cause from symptom\u2014and the genetic tools meant to overcome this limitation inherit a different fatal flaw. A value like a 14% increase in all-cause mortality doesn\u2019t give us an actionable insight, a metric, a goalpost to aim towards. But that failure points us somewhere more useful: Short-term experiments give us clean, causal reads on biology that fundamentally matter for health and longevity.<\/p>\n
Which leaves us with two things worth holding in mind. First, this particular study didn\u2019t teach us much, and it certainly can\u2019t hand you a personal risk figure. But its very limitations help sharpen our aim. Sleep is fundamentally important\u2014and because we know precisely which systems it moves, and in which direction, that knowledge should guide how seriously we protect a good night\u2019s rest. Not because a cohort study assigned us a number, but because we can see, night to night, exactly what\u2019s at stake.<\/p>\n
Second, short term experiments are usually brief and severe\u2014four or five hours of sleep per night, or none at all, for a few days. That is not the occasional late night or the odd hour lost; it\u2019s a far heavier insult. And the systems it harms rebound quickly\u2014if we return to a normal, healthy sleeping pattern.<\/p>\n
So a rough night here and there is nothing to agonize over. In other words, don\u2019t lose sleep over losing sleep now and then. But we should be informed by what the good experiments do show us\u2014sleep loss meaningfully drags on much more than our energy levels the following day\u2014and know that consistently protecting our sleep means protecting fundamental systems driving our health and longevity.<\/p>\n
For a list of all previous weekly emails, click here<\/a>.\u00a0<\/p>\n podcast<\/a> | website<\/a> | ama<\/a><\/p>\n References<\/p>\n 1. Cappuccio FP, Miller MA. Sleep and mortality: cause, consequence, or symptom? Sleep Med. 2013;14(7):587-588. doi:10.1016\/j.sleep.2013.04.001<\/a><\/p>\n 2. Ungvari Z, Fekete M, Varga P, et al. Imbalanced sleep increases mortality risk by 14-34%: a meta-analysis. GeroScience. 2025;47(3):4545-4566. doi:10.1007\/s11357-025-01592-y<\/a><\/p>\n 3. Shi G, Xing L, Wu D, et al. A rare mutation of \u03b21-adrenergic receptor affects sleep\/wake behaviors. Neuron. 2019;103(6):1044-1055.e7. doi:10.1016\/j.neuron.2019.07.026<\/a><\/p>\n 4. Ai S, Zhang J, Zhao G, et al. Causal associations of short and long sleep durations with 12 cardiovascular diseases: linear and nonlinear Mendelian randomization analyses in UK Biobank. Eur Heart J. 2021;42(34):3349-3357. doi:10.1093\/eurheartj\/ehab170<\/a><\/p>\n 5. Lamon S, Morabito A, Arentson-Lantz E, et al. The effect of acute sleep deprivation on skeletal muscle protein synthesis and the hormonal environment. Physiol Rep. 2021;9(1):e14660. doi:10.14814\/phy2.14660<\/a><\/p>\n 6. Nedeltcheva AV, Kilkus JM, Imperial J, Schoeller DA, Penev PD. Insufficient sleep undermines dietary efforts to reduce adiposity. Ann Intern Med. 2010;153(7):435-441. doi:10.7326\/0003-4819-153-7-201010050-00006<\/a><\/p>\n