A steadier clock tracks with less death — untested as a lever
You can change this. Whether tightening the clock itself cuts death is untested — nobody here has shown that making sleep more regular cuts death.
A five-level self-report mapped onto accelerometer SRI quintiles. It is not an SRI score. The 0.70 is fully adjusted most-regular versus least. The 0.52 is the same contrast in the minimal model — not used.
Dose
Very Consistent on this calculator’s five-level self-report. That label is mapped onto Windred’s most-regular SRI quintile (80–100th vs 0–20th = fully adjusted ACM HR 0.70). Interiors 0.91 / 0.84 / 0.76 are log-linear fill, not Table 2. Least-regular is the paper’s own comparator (HR 1.0), not a population typical.
Interiors 0.91 / 0.84 / 0.76 are log-linear fill between 1.0 and 0.70 — constructed, not read off Table 2.
Dose-response shape: monotone
How to get there
Hold the same bedtime and wake time daily — regularity is a behavioral anchor, not a duration change.
Same five-level self-report this page scores, tracked over weeks.
Cribb's reanalysis plateaus past the median — above-median regularity buys much less than the quintile contrast implies.
Evidence, by endpoint
| Endpoint | Grade | Citation |
|---|---|---|
| All-cause mortality, most- vs least-regular SRI quintile | C | Windred et al. 2024, Sleep Engine paper. UK Biobank accelerometer SRI, n=60,977 (mean age 62.8, 55% female, median SRI 81.0), 1,859 deaths, mean follow-up 6.30 years. Least-regular quintile is the comparator (HR 1.0). Most-regular (80–100th) vs least (0–20th): minimal-model ACM HR 0.52 (0.45–0.60); fully adjusted HR 0.70 (0.59–0.83). The engine uses 0.70, not 0.52. Abstract 20%–48% lower ACM across the top four quintiles includes that 0.52. Stronger ACM predictor than sleep duration in that cohort. |
| All-cause mortality across mixed regularity metrics | C | Kalkanis et al. 2025, Sleep Medicine Reviews Best available review of the construct, not a meta. Search to 1 July 2025, 59 primary studies. Heterogeneity of metrics and outcomes precluded meta-analysis. Five low-bias cohorts: 20–88% higher ACM for the least regular sleepers, independent of duration and quality — a range across mixed metrics, not a pooled HR, not the engine number. |
| All-cause mortality vs the median SRI, same UK Biobank wave | C | Cribb et al. 2023, eLife Same UK Biobank wave, different SRI pipeline (GGIR; median SRI 60 vs Windred’s 81) and different comparator (the median, not the least-regular quintile). n=88,975, 3,010 deaths, median 7.1 years. Vs median: 5th percentile (SRI=41) ACM HR 1.53 (1.41–1.66); 95th (SRI=75) 0.90 (0.81–1.00). Non-linear; plateaus past the median. Not the engine paper. |
| All-cause mortality, self-reported irregular vs regular | C | Omichi et al. 2022, Sleep Health J-MICC, n=81,382, mean follow-up 9.1 years. Self-reported sleep/wake regularity — closer to what this calculator asks than SRI. Irregular vs regular ACM HR 1.30 (1.18–1.44), regardless of duration. Binary, not quintiles. Not the engine paper. |
| All-cause mortality, regular-optimal vs irregular-insufficient clusters | C | Chung et al. 2023, Journal of Sleep Research MESA Sleep, n=1,759, median 7.0 years, 176 deaths. Actigraphy clusters mixing regularity and duration: “regular-optimal” (7.0±1.0 h, regular) vs “irregular-insufficient” (5.8±1.4 h, twice the irregularity) ACM HR 0.61 (0.45–0.83). Not an SRI-quintile contrast. Not the engine paper. |
| All-cause mortality from a randomized regularity intervention | X | No trial in this set assigned a sleep-timing regularity target and measured all-cause mortality. Windred compared equivalent models; it did not randomize clocks. |
What the dose is
Very Consistent on a five-level self-report: Very Irregular, Irregular, Somewhat Consistent, Consistent, Very Consistent. Those labels are mapped onto Windred et al. 2024 (Sleep) Sleep Regularity Index quintiles from a UK Biobank accelerometer. The target is the most-regular quintile. The engine prints fully adjusted ACM HR 0.70 (0.59–0.83) there versus the least-regular quintile. It does not print 0.52, and it does not compute an SRI from a watch.
What the evidence shows
Windred is one prospective cohort (n=60,977, 1,859 deaths, mean follow-up 6.30 years), grade C here — not a meta. In that cohort, regularity was a stronger all-cause predictor than duration. Kalkanis et al. 2025 (Sleep Medicine Reviews) is the construct review: 59 studies, no pooled HR, five low-bias cohorts narratively 20–88% higher ACM for the least regular, independent of duration and quality. Omichi et al. 2022 (Sleep Health) is the self-report companion, HR 1.30 (1.18–1.44) irregular vs regular, independent of duration — closer to what this page asks than an SRI.
What the number is not
It is not a demonstration that regularising a clock lengthens life. No trial in this set assigned a regularity target and counted deaths, which is why the intervention endpoint is grade X. It is not Cribb et al. 2023’s 1.53 / 0.90 versus the median, and it is not Chung et al. 2023’s 0.61 for a cluster that also changes hours. It is not the abstract’s 20%–48% range.
How the five levels get scored
Least-regular = HR 1.0 because that is Windred’s comparator, not because it is typical. Interiors 0.91 / 0.84 / 0.76 are log-linear between 1.0 and 0.70 — constructed, named on /methodology, not read off Table 2. Cribb’s plateau past the median is the shape argument against treating those interiors as a smooth climb all the way to 0.70.
What argues against this
Cribb et al. 2023 reanalyzed the same UK Biobank accelerometer wave with a different SRI pipeline and a different comparator. Versus the median, the 5th percentile is HR 1.53 (1.41–1.66) and the 95th is only 0.90 (0.81–1.00), plateauing past the median — a flatter payoff for above-median regularity than Windred’s quintile contrast of 0.70. There is no matching-construct meta: Kalkanis et al. 2025 precluded pooling. Windred is one UK Biobank cohort (mean age 62.8). The 0.52 on older pages is the minimal model at the same contrast, not the engine. Nobody here has shown that making sleep more regular cuts death.
Gotchas
- 0.70 is fully adjusted most- vs least-regular. 0.52 is the minimal model. The abstract’s 20%–48% range includes 0.52. Do not weld it onto the engine.
- The calculator does not compute an SRI. It asks a five-level self-report and maps the labels onto Windred’s quintiles. Omichi’s 1.30 is the closer instrument and a different contrast (binary, not quintiles).
- Cribb’s median SRI is 60; Windred’s is 81. Same wave, different pipeline. 1.53 is not 1 / 0.70.
- Interiors 0.91 / 0.84 / 0.76 are log-linear fill between 1.0 and 0.70, not Table 2.
- Chung’s 0.61 mixes 7.0 h regular with 5.8 h irregular. That is not this slider.
Safety
- This is a timing habit, not a licence to shorten sleep. Duration is scored separately and is U-shaped. Do not crash a night to protect a clock.
Educational content only — not medical advice, diagnosis, or treatment. Not a medical device; not FDA evaluated. Consult a qualified healthcare professional about your own health, and call emergency services for urgent symptoms. Full medical disclaimer →