A facet of the sleep cluster (nucleus Sleep Duration and Mortality) that opens a new exposure
axis. The duration literature asks how many hours; this asks how consistent the timing, and the
two are not the same question. Sleep has at least three separable dimensions — duration (hours),
regularity (day-to-day consistency of sleep-wake timing), and within-night continuity
(fragmentation) — and a person can score well on one and badly on another.
(inferred from Windred et al., 2023)
The finding — regularity out-predicts duration in the same cohort [EXTRACTED]
Windred measured both dimensions objectively (wrist accelerometry, not self-report) in 60 977 UK Biobank participants over >10 million hours, and compared them head-to-head for mortality.
- More regular sleep -> 20-48% lower all-cause mortality across the top four Sleep Regularity Index (SRI) quintiles vs the least-regular quintile (16-39% cancer, 22-57% cardiometabolic). (Windred et al., 2023)
- Regularity was the stronger predictor. Fully-adjusted top-quintile HR: SRI 0.70 (0.59-0.83) vs duration 0.76 (0.65-0.89); model comparison (AIC) favoured SRI (full-model p=.005). Decisively, adding duration to an SRI model did not improve fit — «sleep duration does not explain significant additional variance in mortality risk beyond the variance explained by SRI scores» (nested LR test, full-model p=.20). (Windred et al., 2023)
- The SRI-mortality relationship is monotonic (more regular = lower risk, no U-turn), unlike the duration U-curve. (Windred et al., 2023)
Because SRI and duration were compared in one cohort with one design, this is a within-source comparison — no cross-source commensurability question arises.
Why it does not contradict the duration nucleus — a reconciliation, not a tension
Windred’s duration arm looks weaker and non-U-shaped here, which could read as a clash with Sleep Duration and Mortality (Cappuccio’s strong U-curve). It is not — the two measured different exposures over different ranges, and Windred says so:
- Cappuccio pooled self-reported duration with long-arm cutoffs of >9-10 h; Windred’s longest objective quintile was only >7.56 h — «we therefore would not necessarily expect to see a heightened risk of mortality in this upper quintile». So Windred simply does not reach the long-sleep range where the illness-marker arm lives. (Windred et al., 2023)
- Windred confirms the short-sleep and cardiometabolic-duration associations. It does not overturn the duration finding; it adds an axis the duration literature omitted (type-F refinement of the duration-centric public-health framing). Duration was «the central focus of current sleep health guidelines», yet is here the weaker predictor — while regularity, which guidelines neglect, is the stronger one. (inferred from Windred et al., 2023)
Mechanism — circadian disruption, not a sleep-quantity pathway [INFERRED-directional]
SRI is proposed as a proxy for circadian disruption: irregular sleep-wake timing scatters the timing of light, meals and activity, desynchronizing central + peripheral clocks. This is a different mechanism family from the short-sleep leptin/ghrelin/glucose story on Sleep and Metabolic Health — it is about when, not how much. The cancer signal (irregular sleep predicted cancer mortality; short duration did not, robust in the cancer-free) fits a circadian-oncogenesis pathway. Held directionally, marked as mechanism: the study is correlational and cannot fix causation. (inferred from Windred et al., 2023)
Experimental-animal backing for the circadian-oncogenesis mechanism — directional, and NOT independent corroboration. The circadian-disruption route Windred proposes has experimental-animal support: IARC’s Monographs Working Group rested its Group-2A shift-work classification partly on sufficient animal evidence for carcinogenicity of light during the biological night, via melatonin suppression and clock-gene deregulation -> Night Shift Work and Breast Cancer (where that evidence is held and quoted). This is deliberately NOT logged as type-E independent backing: a human circadian-epidemiology account and the experimental circadian-oncology literature draw on the same antecedent mechanism base, so their agreement is a shared root, not two separately-arrived routes (the laundered-E trap — independence was asserted, never verified). Held as directional mechanism only, and its realized potency on a patient-important human outcome is bounded low: the best-powered human test of this route (night shift work -> breast cancer) is a well-powered NULL.
Decision relevance
- A concrete, low-cost target that is easier than extending sleep. Top-20% SRI = falling asleep and waking within ~1-hour windows most days; bottom-20% = ~3-hour windows. Regularity «may also be an easier dimension to target through interventions» than adding hours (which is psychosocially and biologically hard). (Windred et al., 2023)
- Ranking (layer 1). A candidate moderate lever for the already-adequate-duration stratum — the person sleeping ~7 h but at chaotic times (shift-adjacent schedules, social jetlag) has a lever here that the duration advice misses. Certainty is capped by the single-cohort, correlational design, so it ranks below the established big rocks and is not yet a confident recommendation.
- Not a substitute for the short-sleep lever. Regularity being the stronger predictor does not license neglecting chronic short sleep — both are actionable; regularity is the newly-visible one.
Limits
- Single high-tier cohort, correlational — «Sleep regularity may be both a cause and marker of premature mortality risk». No RCT that raises SRI and measures mortality exists; the intervention claim is inferred from the association + a plausible circadian mechanism, not demonstrated. (Windred et al., 2023)
- One 7-day snapshot; older, 97%-white cohort — transportability and temporal-stability untested; fully-adjusted covariates may be partial mediators, so the true effect likely sits between the minimal (HR 0.52) and full (HR 0.70) models.
- Coherence, not validity (R1): SRI predicts mortality in this cohort; that regularity causes lower mortality, and that raising it would help, are the plausible-but-unproven step.
- AWAITS a second regularity cohort or an SRI-raising trial, and a guideline that targets regularity (current guidelines target duration).
(inferred from Windred et al., 2023)