The U-curve: What the 10–11pm Finding Really Shows

One study put a number on everyone's bedtime anxiety: sleep onset between 10 and 11pm was associated with the lowest cardiovascular risk in the UK Biobank. This page reads that curve the way an epidemiologist would — what it actually shows, what it can't prove, and why your biology matters more than the clock on the wall.

The Chart Everyone Saw

The finding behind the headlines: an analysis of the UK Biobank's accelerometer data linked sleep onset timing to cardiovascular disease incidence in roughly 88,000 people followed for about six years (Nikbakhtian et al., 2021). Sleep onset between 10:00 and 10:59pm sat at the bottom of the curve. Onset at or after midnight was associated with roughly a quarter more incident cardiovascular disease, a statistically significant elevation. Onset before 10pm showed a smaller elevation that did not reach statistical significance, and 11pm–midnight sat in between, also short of significance. The shape — higher risk on both sides of a mid-evening trough — is where the "U-curve" name comes from.

The press version of this study was predictable: "10pm is the perfect bedtime," rendered as a rule with the caveats stripped off. The honest version is flatter and more useful: the data mark very late onset as the group to take seriously, and they give no license to frighten early sleepers or to bully owls toward a clock hour their biology will not honor. Population curves describe where groups land; they were never designed to prescribe where you should.

The Shape of the Curve

Look closely at what the curve does not say. The dramatic arm is the late one: midnight-and-beyond carries the significant elevation. The early arm — before 10pm — is a modest bump that could easily be chance. Yet the early arm is the one that frightened a certain kind of reader: people who already sleep at 9:30pm wondering if their disciplined bedtime is quietly dangerous. It is not, at least not on this evidence. The honest summary of the shape: very late is the risk; mildly early is a shrug.

What the U-curve Can't Prove

The study is observational, and the groups differ in ways that matter enormously. People sleeping after midnight are younger on average, more often unemployed or shift workers, and differently composed in health status than the 10pm group — and while the analysis adjusted for many of these, adjustment is a statistical patch, not a time machine. The deeper problem runs both directions:

One confounder deserves naming: reverse causation. Illness makes people sleep earlier — fatigue, pain, and medication all drag bedtime backward — so some share of the early arm's elevation may be sick people migrating toward 9pm, not 9pm making people sick. The same logic means the late arm is partly life structure: night work, unemployment, and chaotic schedules travel with late bedtimes, and each carries its own health burden independent of the hour.

Why Individual Timing Matters More

The strongest reason to distrust any universal bedtime is what the curve cannot see: the person's chronotype. The risk in these cohorts appears to track misalignment — sleep forced outside the window your biology sets — rather than the clock hour itself. The same midnight onset that marks the high-risk group for a lark forced into late shifts is an owl's ordinary, well-matched bedtime. The chronotypes page and social jetlag page cover the evidence that mismatch, not lateness per se, does the metabolic damage.

That said, timing is not nothing. A Mendelian randomization study — using genetic variants that nudge people toward earlier sleep timing, which sidesteps the usual confounders — found earlier timing associated with lower depression risk (Daghlas et al., 2021). This is one of the few pieces of evidence that sleep timing itself, not just the behaviors around it, can carry causal weight. But the direction is individual: the genetics shift phase relative to a person's own rhythm. It is evidence for aligning with your biology, not for a universal lights-out.

Consistency Outranks the Clock

If you want a timing finding with more robust support than the U-curve, it is regularity. A UK Biobank analysis found that the regularity of sleep timing — quantified as a sleep regularity index — was associated with mortality risk more strongly than sleep duration itself (Windred et al., 2024). A systematic review of sleep timing and consistency reached a compatible conclusion: consistency of the sleep–wake schedule is associated with better health outcomes, and the evidence for consistency is at least as strong as the evidence for any particular hour (Chaput et al., 2020). Translated: a "wrong" time kept steadily likely beats a "right" time kept erratically — which is exactly what the parent topic flagged and this page now puts in context.

So What Time Should You Sleep?

The synthesis the evidence actually supports, in order of confidence:

• 1️⃣ Compute your own window. The formula — wake time minus need minus buffer — respects your life; the clock hour does not.
• 2️⃣ Match your chronotype. A consistent 12:30am lights-out that fits an owl's phase is a different animal from a lark forced to midnight (see the chronotype page).
• 3️⃣ Protect regularity first. Same window ±1 hour, weekends included — the habit with the most consistent support in this literature.
• 4️⃣ Treat 10–11pm as a default, not a law. If your computed window lands near mid-evening anyway — as it does for most lark-and-middle adults — the data agree with you. If it doesn't, the data about other people do not outrank your phase.

Questions, Answered Briefly

• 📊 Should I move my bedtime to 10:30pm? Only if that fits your computed window and chronotype. Moving against your phase to chase a cohort average trades a speculative benefit for a certain mismatch.
• 🌅 Is my 9:30pm bedtime dangerous? On this evidence, no — the early arm's elevation was small and not statistically significant, and plausible confounders (illness, age) point the other way.
• 🌙 Is my 1am bedtime dangerous? It sits in the group with the significant risk elevation — but how much of that is your hour versus your mismatch or your schedule is exactly what the study cannot say. If 1am fits your phase and your life, protect regularity; if it doesn't, shift.
• 🔬 Will a trial ever settle this? A randomized bedtime trial is logistically brutal, which is why the field leans on cohorts, natural experiments, and genetics. Expect refinement, not a verdict.

Related Topics