Ninety minutes of delayed bedtime. Seventy-eight point four minutes of actual sleep lost. One pound gained in six weeks.
Those are the numbers from Marie-Pierre St-Onge's lab at Columbia, published this month in the Annals of Internal Medicine, and they are the most useful numbers in sleep science right now precisely because they are so small.
Last week in this space we covered the rest-state itself — sleep inertia on waking, the dietary compounds people reach for to fall asleep faster. This week we move to a harder question: how much sleep, and in which direction, actually moves the needle on health — and whether the scary long-sleep numbers everyone quotes mean anything at all.
Start with the Columbia trial, because it is the rare piece of sleep evidence that isn't an observational correlation. St-Onge recruited 95 healthy adults who normally slept seven to eight hours. Each person did two six-week phases: one on their usual schedule, one with bedtime pushed back about 90 minutes.
That design is the point. Earlier sleep-loss experiments starved volunteers down to four hours a night, but those studies collapsed after a few days because almost nobody can tolerate it. St-Onge instead reproduced the deficit millions of people actually live with — an hour or so short, sustained for weeks.
The result: about one pound over six weeks, plus a wider waist, in a randomized crossover where each subject served as their own control.
One pound sounds trivial. St-Onge's framing is that it isn't, if it doesn't stop. "The main takeaway is that sustaining a reduction in sleep, to a duration that is often observed in the general population, leads to weight gain and increases in waist circumference," she told Medical News Today. Projected across a year, the researchers argue, this crosses into clinically significant territory.
The mechanism is more interesting than the pound. Participants didn't just eat more — they moved less, logging about 17 extra sedentary minutes a day. Among men and postmenopausal women, inactivity rose by nearly 30 minutes.
And that sedentary bump held even after accounting for the extra waking hours the short sleepers had available. Being awake longer wasn't the story. Feeling too tired to move was.
"Weight gain occurs through an imbalance in energy expenditure and energy intake," St-Onge explained. "We have shown in previous studies that participants eat more when sleep is restricted. I think both come into play." Her subjects were screened to be relatively inactive to begin with, so short sleep worsened an already weak lifestyle factor rather than creating one.
Take the evidence level seriously here. This is a randomized crossover trial — strong for causation — but small, six weeks long, and unable to tell us what happens over years. The team says so plainly.
When the bigger effect size is the one with no mechanism behind it, that's a flare signaling the whole signal is tangled with underlying illness.
Around that solid core sits a large, much shakier literature on sleep duration and death. And this is where the story gets genuinely contested.
The headline number circulating in longevity spaces comes from a meta-analysis of 79 observational studies: sleeping fewer than seven hours a night is associated with a 14 percent increase in all-cause mortality. Reasonable-sounding. Widely repeated as though sleep loss causes the deaths.
The tell is buried in the same data. Long sleepers — nine hours or more — showed a 34 percent jump in mortality, more than double the short-sleep effect.
Here's the problem, as Peter Attia lays it out. There is no plausible biological mechanism by which simply sleeping an extra hour shortens your life. The causal evidence base runs one direction — sleep loss harms — not sleep surplus.
When the bigger effect size is the one with no mechanism behind it, that's a flare signaling the whole signal is tangled with underlying illness. The simpler read: long sleep is a symptom. The sick, the frail, and the not-yet-diagnosed sleep more.
This is reverse causation, the trap that haunts every observational sleep study. Sleep can be a cause of disease, a consequence of it, and a symptom of it — sometimes all three in the same person. Heart failure, cancer, and depression all perturb sleep long before diagnosis, and no amount of statistical adjustment fully clears that out.
The supposed fix is Mendelian randomization, which uses gene variants fixed at conception as a natural randomized experiment. Run it, and genetically predicted short sleep does predict cardiovascular disease — hypertension, heart attack, coronary disease. The long-sleep association mostly evaporates, exactly as the "symptom, not cause" reading predicts.
So far, so good. But MR only works if the sleep genes affect disease only through sleep. For the brain, that's rarely true.
Attia's example is ADRB1. A variant in this gene shortens sleep by making wake-promoting brainstem neurons easier to fire. But ADRB1 also encodes the β1-adrenergic receptor on heart muscle — the exact target of the beta-blockers cardiologists prescribe daily.
So a genetic link between short sleep and heart disease through ADRB1 could just be the receptor acting directly on the heart. Sleep loss may be riding along, not driving. A recent MR study leaned on 78 genes, most poorly understood, to isolate its effect — and nobody has dissected them one by one to rule this out.
A 2013 critical review of 42 prospective studies reached a quieter version of the same conclusion: it is premature to claim a robust U-shaped curve between sleep and mortality across populations. Most studies used self-reported sleep, which correlates poorly with what wrist devices actually measure.
Which is why the two most credible new signals on long sleep are worth watching — not because they prove harm, but because they hint at biology.
A UT Health San Antonio study of 2,410 Framingham participants, average age 70, found that sleeping eight-and-a-half to nine hours or more was associated with higher blood levels of p-tau181, an Alzheimer's-linked protein — the association climbing sharply beyond ten hours, after adjusting for age, sex, sleep apnea, depression, kidney function, and APOE ε4 status.
Lead author Vanessa Young is careful. "Because this is a snapshot in time rather than a long-term study, we cannot say that long sleep causes Alzheimer's, but the findings suggest it may be worth monitoring, and that more sleep is not always better for brain health."
Notably, three other neurodegeneration markers lost their sleep association once kidney function was accounted for. Only p-tau181 survived. That selectivity is what makes it interesting rather than noise.
A 7-year Chinese cohort of 7,984 adults over 45 pushed the direction question harder using cross-lagged models. Both short and long sleep predicted worse later cognition — and worse cognition predicted abnormal sleep more strongly than the reverse.
Their conclusion: cognition was the major driver of the loop. In older brains, in other words, the arrow points at least as much from illness to sleep as from sleep to illness — precisely the reverse-causation Attia warns about, caught in the act.
So where does the evidence actually leave you.
Mild chronic short sleep — the hour-plus deficit most people run — has causal, trial-grade support for gradual weight gain and reduced movement. That threshold is real and worth defending.
The claim that sleeping more than eight or nine hours will shorten your life does not have that support. The mortality and biomarker signals at the long end are more consistent with sickness reaching backward into sleep than with sleep reaching forward into disease.
If you naturally sleep a little under eight, protect it — the metabolism data says so. If you find yourself needing nine or ten hours to function, the more useful question isn't how to sleep less. It's what your body is telling you it's fighting.