Skip to content
What Causes What
SMOKE~ NON-LINEAR

Sleep duration and regularity → All-cause mortality

Higher Sleep duration and regularity moves All-cause mortality in a non-linear way. Dose or timing can flip the sign.

Yin 2018's dose-response meta of prospective cohorts found a U-shaped link between hours of sleep and all-cause death: each hour under 7 tracked a 6 percent higher risk, and each hour over 7 tracked 13 percent. Nobody randomized time in bed.

Meta-analysis · Low · N=2200425

The association disappears when you account for other factors, like socioeconomic status or healthy-user bias.

Mechanism

Hours asleep and how regular those hours are. Death from any cause. The duration metas are U-shaped or J-shaped. Both short and long sleep track more death. Yin, Jin, Shan and colleagues (2018) pooled prospective cohorts to December 2016 in generally healthy people. Lowest all-cause death sat near 7 hours a day. Below 7 hours the pooled RR was 1.06 (1.04 to 1.07) per hour cut. Above 7 it was 1.13 (1.11 to 1.15) per hour added. Liu, Xu, Rota and colleagues (2018) fit a restricted cubic spline on 40 cohorts, 2,200,425 adults, 271,507 deaths. Versus 7 hours, 4 hours RR 1.05, 6 hours 1.04, 9 hours 1.13, 10 hours 1.25, 11 hours 1.38. Nonlinear P less than 0.01. Cappuccio, D'Elia, Strazzullo, and Miller (2010) pooled 16 studies, 27 samples, 1,382,999 people, 112,566 deaths. Short sleep RR 1.12 (1.06 to 1.18). Long sleep RR 1.30 (1.22 to 1.38). Questionnaire sleep. Certificate death. Windred, Burns, Lane and colleagues (2024) scored sleep regularity from more than 10 million accelerometer hours in 60,977 UK Biobank adults. 1,859 deaths, mean 6.3 years. The top four regularity fifths tracked 20 to 48 percent lower all-cause death than the least regular fifth. Regularity beat duration in nested models. the record here is The paper names N=60,977. Liang, Ai, Xue and colleagues (2023) followed 90,398 of that accelerometer wave. Short sleep HR 1.27 (1.11 to 1.45). Long sleep HR 1.16 (1.06 to 1.28). Long sleep plus lower efficiency HR 2.11 (1.44 to 3.09). Liu, Niu, Duan and colleagues (2023) followed 5,772 older Chinese adults (median age 82). A 3-year drop of more than 3 hours HR 1.26. Short to long 1.28. Long to short 1.52. Liang, Feng, Chen and colleagues (2023) followed 92,221 of the same UK Biobank wave. Without recommended activity, short sleep HR 1.88 and long sleep HR 1.69 versus recommended activity and normal sleep. Authors name metabolism, immune load, and illness that changes sleep. The files do not isolate one path.

Caveats

The stamp is smoke because nobody randomized hours in bed. These are cohorts and metas of cohorts. Observation alone cannot close the backdoor, so the stamp stays smoke. More papers of the same design do not settle the claim. A larger meta is still observation. Reverse causation sits on the backdoor. Illness can shorten sleep, lengthen it, or wreck the clock, and also raise later death. Healthy-user habits travel with a regular 7-hour night: income, movement, less drinking. Residual confounding remains. Yin's cardiovascular and stroke curves are not this page. This page is all-cause death. Sleep apnea is a different question. Cappuccio's sleep was a questionnaire. Later UK Biobank files use a wrist accelerometer. Windred, Liang 90,398, and Liang 92,221 share that wave. They are not three independent samples. This is a journal entry, not medical advice and it is not a reason to pick a number of hours.

Effect

qualitative. Yin 2018: U-shape, nadir ~7h; <7h RR 1.06/h (1.04-1.07), >7h RR 1.13/h (1.11-1.15). Liu 2018: 40 cohorts, 2,200,425, J-shape. Both arms raise death. Design does not identify.

Nodes

Lifestyle & environmentSleep duration and regularityBiology & endocrinologyAll-cause mortality

Named confounders

Cite this page

What Causes What. “Sleep duration and regularity → All-cause mortality.” https://whatcauseswhat.org/edges/e-sleep-mortality (atlas updated 2026-09-02).

https://whatcauseswhat.org/edges/e-sleep-mortality

Papers