The Number Everyone Chases
Ask how someone slept and you get a number. Sleep trackers score the night in hours and minutes, guidelines draw their lines the same way, and the whole conversation treats sleep like a bank account where deposited hours collect a health dividend. The logic feels airtight: short sleep predicts early death across dozens of cohorts, very long sleep does too, and the famous U-shaped curve makes the advice write itself. The question is whether we have spent thirty years optimizing the wrong dial.
The Study That Timed the Clock
In 2024 the journal Sleep published this bluntly titled paper: "Sleep regularity is a stronger predictor of mortality risk than sleep duration." Daniel Windred, Andrew Phillips, and colleagues at Monash University stopped asking how long 60,977 UK Biobank adults slept and measured something else: how regularly they did it, via wrist accelerometers worn for seven straight days that generated more than 10 million hours of movement data. From those traces the team computed a Sleep Regularity Index, scored from -100 to 100, capturing the probability of being in the same state at the same clock time on consecutive days; 100 means an identical daily schedule, and the median scored 81. Over 6.3 years, 1,859 participants died; after adjusting for age, sex, ethnicity, socioeconomic position, lifestyle, and baseline health, the four more regular quintiles carried 20 to 48 percent lower all-cause mortality than the least regular fifth, with cardiometabolic death 22 to 57 percent lower, cancer death 16 to 39 percent lower. The direct comparison was the point: in nested models, adding sleep duration to a model already containing regularity improved prediction by nothing (p = 0.14 to 0.20); regularity carried the signal while duration, once regularity was known, added noise.
The Weekend Math
The paper never runs this calculation, though the formula allows it: the index compares each minute against the same minute 24 hours later, so every shifted hour creates roughly two hours of mismatched minutes, registering against both neighboring days. Take an ordinary pattern, 11 p.m. to 6 a.m. on weeknights sliding to 12:30 a.m. to 7:30 a.m. on Friday and Saturday: that 90-minute shift on two nights produces about 360 mismatched minutes out of 8,640 weekly comparisons, and the formula (SRI = -100 + 200 x (1 - 360/8,640)) lands at roughly 92, about 8 points below perfect. The study's interquartile range spanned just 12.5 points (73.8 to 86.3), so two late weekend nights erase roughly two-thirds of the distance between a 25th- and 75th-percentile sleeper; the social jet lag most of us consider harmless is most of the gap between average and excellent. Two caveats: the arithmetic assumes a clean week with exactly two shifted nights, while real weeks are messier (pushing scores down, not up), and the index is symmetric, so shifting earlier costs the same as later.
It Replicated
One cohort, however large, is only a hint. The finding has now survived two independent re-analyses: first, Lachlan Cribb's team re-analyzed 88,975 UK Biobank participants with different exclusions and adjustments (eLife, 2023) and the shape matched, with 53 percent higher mortality hazard at the 5th percentile of regularity versus the median and cardiovascular and cancer mortality following the same curve. Chaput's group then pushed the question into heart disease: 72,269 adults tracked for eight years in the Journal of Epidemiology and Community Health (2024), with irregular sleepers at 26 percent higher risk of major cardiovascular events (heart attack, heart failure, stroke). The sharpest detail came from the joint analysis: among moderately irregular sleepers, getting the recommended hours erased the excess risk, but among the truly irregular it did not. You cannot out-sleep a chaotic clock.
The Strongest Counterargument
The best case against this story is that irregular sleep is a symptom, not a cause: erratic schedules travel with shift work, chronic pain, depression, a newborn, a dying parent, poverty, and thin apartment walls, each of which raises mortality on its own. The authors adjusted for a long roster of confounders, yet adjustment is not randomization, and no regression fully captures what rotating nights do to a body. Reverse causation sharpens the objection: failing health scrambles sleep long before a diagnosis enters a medical record, and one seven-day snapshot of wrist movement cannot establish which came first, the irregularity or the illness behind it. The eLife replication excluded people with baseline disease and still found the pattern, which weakens this challenge without defeating it. The cohort sets its own limits: mean age 63, 97 percent white, volunteers healthier and wealthier than the population at large, so whether regularity outranks duration for a 30-year-old nurse, a teenager, or anyone outside Britain is untested. And the decisive experiment does not exist: nobody has randomized people to regular versus irregular sleep and counted the deaths, so until that trial runs, "fix your schedule, live longer" remains an inference, not a prescription. Duration, finally, did not stop mattering: the minimally adjusted models kept its U-shape, so regularity won the head-to-head without licensing four steady hours over eight steady ones.
What We Didn't Prove
No study here proves that adopting a regular schedule extends your life: the exposure was measured once, by wrist movement, across one week between 2013 and 2016, while habits drift, devices misclassify stillness and naps, and deaths accumulated years later. Both cohorts skew old, white, and British, and the eLife replication drew from the same UK Biobank well: an independent analysis, not an independent population. A randomized trial of sleep regularity, the experiment that would settle causation, has never been run.
The Bottom Line
Thirty years of sleep advice optimized a number: the hours. The best device-measured evidence now says the clock matters more than the count, because across 60,977 adults, the consistency of sleep timing predicted who died more strongly than the amount of sleep, and two independent follow-ups agreed. Regularity is also the cheaper fix: adding an hour of sleep is hard, while keeping the hour you already have is a decision.
What You Can Do
Anchor your wake time, weekends included: the circadian system takes its strongest cue from when you rise and meet daylight, so a fixed alarm plus morning light beats a bedtime you cannot keep, and when one end of the night has to give, protect the morning. Keep the window tight: the most regular fifth held sleep and wake times within roughly an hour day to day while the least regular fifth sprawled across three, so aim for the one-hour version and treat a 90-minute weekend slide as what the arithmetic says it is. Track the right metric: many wearables now report schedule consistency or a regularity score next to total hours, so watch that trend instead of the nightly number alone, because a steady six and a half hours beats a swinging eight. Retire the catch-up logic: sleeping until noon on Sunday to repay a short week feels restorative, and by this evidence it partly undoes the repayment, so when the week goes badly, hold the wake time and take a short nap instead, since by the index arithmetic a one-hour nap costs about three points while a three-hour lie-in costs eight.