🧬 Aging Wiki

sleep

Properties1
aliasessleep hygiene, sleep duration, sleep quality, healthy sleep, restorative sleep

20 min read

Sleep (duration, quality, timing)

Sleep is the most universal biological-reset process in mammals, yet it is routinely shortened and degraded in modern populations. In the context of aging biology, sleep is both a diagnostic window (poor sleep quality is an early and sensitive marker of neurodegeneration and systemic inflammaging) and an interventionally tractable target — structured behavioral interventions (CBT-i, light hygiene, chronotype alignment, CPAP for obstructive apnea) demonstrably improve sleep architecture and established clinical outcomes. Lifespan extension from improved sleep has not been demonstrated in a controlled trial. Proposed aging mechanisms include state-dependent brain-fluid transport, HPA-axis regulation, and immune modulation, but their contributions to long-term human outcomes are not established.

Sleep architecture

Normal adult sleep cycles through approximately five 90-minute ultradian cycles per night:

StageAlso calledProportion of sleepFunction
N1Light NREM~5%Transition; brief arousal threshold
N2Core NREM~45–50%Memory consolidation onset; sleep spindles; K-complexes
N3Slow-wave sleep (SWS), deep NREM~15–20% in young adultsDeclarative memory; state-dependent brain-fluid dynamics; GH secretion
REMREM sleep~20–25%Emotional memory; synaptic plasticity; procedural consolidation

Distribution shifts across the night: SWS predominates in the first half of the night (deep restorative sleep is front-loaded); REM predominates in the second half. Truncating sleep from the end disproportionately removes REM; from the beginning (delayed bedtime) disproportionately removes SWS.

Sleep architecture deteriorates measurably from the third decade onward 1:

  • Total sleep time (TST) declines approximately 10 minutes per decade from age 30–60.
  • SWS (N3) undergoes the most dramatic age-related loss: from ~15–20% of sleep in young adults, declining substantially by age 70 — a ~75–80% reduction in slow-wave activity (SWA) amplitude over frontal EEG derivations relative to young adults, with the largest age-related changes during the first NREM sleep cycle 1. SWS time also declines, though the magnitude varies by sex (men show greater loss) and measurement method; some older adults retain more SWS time than others despite severely blunted SWA amplitude.
  • Sleep efficiency (time asleep / time in bed) declines; fragmentation (micro-arousals) increases.
  • Circadian amplitude flattens; the core body-temperature nadir advances (phase advance), shifting preferred bedtime and wake time earlier.
  • REM declines modestly and later than SWS.

These changes are not simply benign “normal aging” — they track strongly with cognitive performance, Alzheimer’s risk, and systemic inflammation, and some are partially reversible with exercise or sleep hygiene intervention.

Glymphatic transport — a contested proteostasis mechanism

The glymphatic-system is a proposed perivascular CSF–interstitial-fluid transport and solute-clearance process associated with astrocytic AQP4. Its stages—CSF influx, parenchymal exchange or redistribution, and net efflux—must be distinguished from one another and from downstream meningeal-lymphatic drainage.2

Foundational mouse evidence: Xie et al. measured a cortical extracellular-space fraction of 14.1% while awake versus 23.4% during natural sleep, a 9.3-percentage-point or approximately 66% relative increase. In a separate terminal cohort, injected amyloid-β1–40 cleared approximately twofold faster and inert inulin more than twofold faster during sleep than wakefulness.3 Tortuosity was unchanged. These experiments used young male mice and acute exogenous tracers; they did not test endogenous protein production, chronic deposition, cognition, aging, or humans.

Why the conclusion is contested: Miao et al. reported lower disappearance of an intraparenchymally injected small dye during sleep and three anesthetic states, despite unchanged local diffusion—the opposite direction from the canonical net-clearance model.4 A 2025 Matters Arising and reply disputed the definition of clearance, modeling, sampling, and analysis without resolving the disagreement.5 Meanwhile, Hauglund et al. found that norepinephrine-linked slow vasomotion drives CSF influx and a clearance readout during natural NREM sleep in mice, and that zolpidem suppresses the relevant oscillations.6 Together, the evidence strongly supports sleep-state changes in brain-fluid dynamics; it does not justify treating tracer influx, oscillation amplitude, and net endogenous-solute removal as equivalent. gap/contradictory-evidence

Human boundary: Human EEG–MRI studies show sleep-coupled CSF and vascular oscillations, and invasive tracer studies provide limited clearance evidence in selected clinical cohorts. No safe non-invasive test currently measures the complete pathway or demonstrates that increasing a glymphatic proxy improves cognition or slows neurodegeneration. DTI-ALPS is especially indirect; see dti-alps. gap/needs-human-replication

DTI-ALPS in ME/CFS: preliminary human association

Thapaliya et al. 2026 compared 31 adults with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) with 27 healthy controls. The bilateral diffusion tensor imaging along the perivascular space (DTI-ALPS) index was lower in ME/CFS (1.44 ± 0.086 vs 1.51 ± 0.11; raw p=0.014, FDR-adjusted p=0.028), with a right-hemisphere group difference but no left-hemisphere difference and no significant between-group asymmetry difference. Among the 29 cases with symptom-item data, the bilateral index correlated with single-item self-reported sleep disturbance (r=−0.47, p=0.013) and impaired concentration (r=−0.43, p=0.026); fatigue and the separate WHODAS cognition score were among the null associations.7

This is disease-specific, cross-sectional evidence—not a sleep intervention or an aging study. DTI-ALPS measures local periventricular directional diffusivity, not waste clearance directly; a separate n=56 comparison with intrathecal contrast-enhanced MRI found limited correspondence and no association with 48-hour tracer dynamics. The defensible interpretation is therefore a lower ALPS index associated with two self-reported symptoms, not demonstrated glymphatic failure or a causal sleep→clearance→cognition pathway.8 See dti-alps for the method-level evidence rubric. gap/needs-replication

Sleep and inflammation — the HPA/NF-kB axis

Poor sleep activates pro-inflammatory signaling at multiple levels 9:

  • HPA-axis: Fragmented or shortened sleep elevates cortisol (and disrupts its normal nocturnal suppression), driving glucocorticoid receptor signaling that paradoxically upregulates NF-kB-driven inflammation over time via receptor desensitization. Sleep is the primary behavioral modulator of the hpa-axis diurnal rhythm.
  • Sympathetic nervous system: Sleep loss increases sympathetic tone (elevated norepinephrine), activating beta-adrenergic signaling on immune cells; this reduces glucocorticoid sensitivity in monocytes and promotes pro-inflammatory transcription.
  • TLR4 pathway: Carroll et al. 2015 examined partial sleep deprivation (restricted to 03:00–07:00 sleep window) in older adults (n=49, ages 60–84) vs younger adults (n=21, ages 25–39) and found a significant age × sleep-loss interaction (p<0.05). Younger adults showed increased TLR-4-stimulated monocyte IL-6 and TNF-α production after partial sleep deprivation; older adults paradoxically showed reduced baseline TLR-4-stimulated cellular inflammation that was not further activated by sleep loss — suggesting that older adults may already be in a state of chronic immune dysregulation that is not acutely amplified by sleep deprivation 9. gap/needs-replication — the blunted acute inflammatory response in older adults does not imply protection; chronic low-grade inflammation in aged monocytes is well-documented separately.
  • IL-6, TNF-alpha, CRP are all elevated in epidemiological studies of chronically short sleepers.

This inflammatory pathway connects sleep quality to chronic-inflammation and likely mediates some of the cardiovascular and metabolic risk associated with short sleep.

Sleep duration and mortality — epidemiological signal

The relationship between sleep duration and all-cause mortality is U-shaped, with lowest risk at ~7 hours per night 10:

Sleep durationAll-cause mortality risk vs 7 hr referenceSource
Short (<7 hr)RR 1.06 (95% CI 1.04–1.07) per 1-hr reduction; pooled RR 1.12 (95% CI 1.06–1.18) for categorical short sleepYin 2017 dose-response; Cappuccio 2010 categorical
7–8 hr/nightReference (lowest risk)Yin 2017
Long (>8 hr)RR 1.13 (95% CI 1.11–1.15) per 1-hr increment; pooled RR 1.30 (95% CI 1.22–1.38) for categorical long sleepYin 2017 dose-response; Cappuccio 2010 categorical

Both meta-analyses report relative risk (RR), not odds ratios. Long sleep (>8–9 hr) shows a stronger mortality association than short sleep; this is partly confounded by reverse causality (underlying disease increasing sleep need), but remains after prospective adjustment in most meta-analyses 11.

Recommended adult sleep duration: 7–9 hours per night (National Sleep Foundation / AAoSM consensus). Both short (<6 hr) and long (>9 hr) are associated with elevated mortality.

Slow-wave sleep and cognitive aging

SWS serves multiple restorative functions that are directly relevant to aging 1:

  1. Memory consolidation: SWS-specific hippocampal replay consolidates declarative memories. SWS loss with age impairs overnight hippocampal-to-cortical memory transfer.
  2. Brain-fluid dynamics: NREM slow waves, vascular-volume changes, and CSF oscillations are coupled in humans; whether SWS maximizes net endogenous-solute clearance remains unresolved.
  3. Growth hormone (GH) secretion: The dominant GH pulse of the 24-hour cycle occurs in early-night SWS; GH decline with age is partly attributable to SWS loss.
  4. Sleep spindle-SWS coupling: Mander et al. (from Walker lab) showed that prefrontal cortical thinning with age impairs NREM slow-oscillation generation, uncoupling slow waves and sleep spindles and predicting overnight memory retention failure — a mechanism connecting brain atrophy to functional cognitive decline via sleep 1.

Sleep-disordered breathing — obstructive sleep apnea

Obstructive sleep apnea (OSA) prevalence rises steeply with age: ~10–17% in adults aged 30–69 (AHI >15), and up to 50%+ in adults >65 by polysomnography depending on diagnostic threshold. OSA causes:

  • Nocturnal hypoxemia (repeated desaturation events)
  • Sleep fragmentation (arousals per hour = AHI events)
  • Suppression of SWS and REM
  • Elevated sympathetic tone and HPA activation

Relevance to aging hallmarks: OSA is an independent risk factor for cardiovascular disease, type 2 diabetes, and cognitive decline — pathways overlapping with hallmarks of aging. Whether OSA directly accelerates biological age (epigenetic clock aging) is under study; some CPAP intervention data show modest improvement in inflammatory biomarkers (IL-6, CRP), but evidence for clock-aging reversal is preliminary. gap/needs-replication

CPAP adherence problem: Despite CPAP being first-line therapy, adherence in elderly populations is poor (~40–60%), limiting population-level benefit.

Circadian misalignment

Circadian misalignment — when sleep timing conflicts with the endogenous circadian clock (e.g., night shift work, social jet lag, late chronotype in early-rising schedules) — produces aging-relevant metabolic effects independent of total sleep duration:

  • Shift workers have elevated risks of metabolic syndrome, cancer (particularly breast and colorectal), and cardiovascular disease, consistent with disrupted circadian regulation of cell-cycle checkpoints, repair processes, and immune surveillance.
  • Social jet lag (>1–2 hr difference between weekday and weekend sleep timing) correlates with higher BMI, elevated CRP, and insulin resistance in population studies — all hallmarks of accelerated aging. gap/needs-replication for hard-endpoint evidence.
  • Chronotype-aligned sleep (sleeping at one’s natural preferred time) is associated with lower cardiometabolic risk than chronotype-misaligned sleep of equal duration.

This mechanism is relevant to altered-intercellular-communication: circadian disruption impairs time-dependent intercellular signaling (e.g., glucocorticoid pulsatility, growth hormone rhythms, immune oscillations) that coordinate tissue-level homeostasis.

Interventional approaches

ApproachTarget populationEvidence levelNotes
CBT-i (Cognitive Behavioral Therapy for Insomnia)Chronic insomnia; any ageStrong — gold standard; outperforms pharmacotherapy long-termComprises sleep restriction therapy, stimulus control, sleep hygiene, relaxation; delivered 4–8 sessions; digital (dCBT-i) formats validated
Sleep hygieneGeneral populationModerate — as standalone, weak; as adjunct, standard of careFixed wake time, darkness, temperature, caffeine cutoff, screen light minimization
Bright light therapyCircadian phase disorders, delayed or advanced sleep phase, shift workers, seasonal moodModerateMorning bright light (>10,000 lux) advances circadian phase; evening avoidance delays it
CPAPOSA (AHI >15 or symptomatic)Strong for symptomatic relief and cardiovascular risk; variable for mortalityAdherence is the rate-limiting factor
Chronotype-aligned schedulingSocial jet lag, shift workEmergingAligning sleep timing to chronotype reduces misalignment burden; employer/school-schedule dependent
Pharmacological hypnoticsShort-term only; not recommended in elderlyCaution — benzodiazepines and Z-drugs associated with excess mortality, fall risk, cognitive impairment, and in the elderly with dementia risk in long-term useNOT recommended as an aging intervention; opposite of restorative

CBT-i is first-line for chronic insomnia by AASM guidelines; pharmacological hypnotics should not be considered as aging-biology interventions.

Biological-age clock effects

The Waziry 2023 CALERIE trial established DunedinPACE as the clock most sensitive to metabolic lifestyle interventions; sleep-focused RCTs with DunedinPACE primary endpoints have not yet been completed 12. Observational data suggest:

  • Short sleep (<6 hr) is associated with accelerated epigenetic aging by multiple clocks in cross-sectional analyses, but causal direction is uncertain.
  • The proposed RCT (sleep extension + CBT-i, 18 months, DunedinPACE primary endpoint in short-sleeping middle-aged adults) would provide the cleanest causal evidence.

See dunedinpace-2022 for clock methodology and cross-intervention comparison. gap/needs-replication — no RCT with biological-age-clock primary endpoint has been completed for sleep interventions as of 2026.

2025–2026 sleep × biological-age-clock cohort and MR evidence

The observational and Mendelian-randomization base for sleep × epigenetic clocks expanded substantially in 2025–2026:

Diao 2025 (Clin Epigenetics, Dongfeng-Tongji cohort, n=3,566 mean age 65.5 yr): Healthy sleep score (composite of bedtime, duration, quality, midday napping) inversely associated with PhenoAgeAccel (β=−0.208, p<0.05), GrimAgeAccel (β=−0.107), DunedinPACE (β=−0.008), and DNAm mortality risk score (β=−0.019). DunedinPACE mediated 6.2% (95% CI 0.8–11.5%) of the inverse association between healthier sleep and reduced all-cause mortality risk over 5.4 years follow-up 13. Stronger associations in older adults than younger (interaction p=0.027 for DunedinPACE). This is the first study to formally quantify the mediation fraction of sleep → mortality through DunedinPACE.

Wu 2025 (Aging (Albany NY), UK Biobank n=442,664): Restricted-cubic-spline + Mendelian-randomization analysis of self-reported sleep duration on PhenoAgeAccel, BioAgeAccel, and leukocyte telomere length 14. Observational analyses confirmed the U-shape (optimum ~7 h/d). MR analyses corroborated the deleterious impact of insufficient sleep but not excessive sleep, suggesting the long-sleep association is largely driven by reverse causation (illness → more sleep), while short sleep has a causal contribution. Cell-type enrichment linked short sleep to BioAgeAccel and LTL through muscle-maintenance and immune-function pathways. This refines the U-shape framing — for short sleep, the causal arrow is supported; for long sleep, residual confounding likely dominates.

Zhang 2026 (Clin Epigenetics, MR for chronotype, napping, sleep duration): Multivariable MR (MVMR) on chronotype, napping, and sleep duration vs telomere length, facial aging, IEAA/HannumAge/PhenoAge/GrimAge, frailty, and cognitive performance 15. Daytime napping had causal independent effects on GrimAge (β=+1.08, p=0.046) and frailty (β=+0.29, p<0.001); longer sleep duration independently protected against frailty (β=−0.36, p<0.001); chronotype (morningness) protected facial aging and cognition after MVMR adjustment. The napping-causal-on-GrimAge finding is novel and adds a sleep-trait dimension beyond duration.

O’Toole 2025 (medRxiv, MULTI study Sleep Chart): Across 23 biological-age clocks spanning 17 organ/tissue systems and 3 omics types, both short (<6 h) and long (>8 h) sleep durations associated with elevated biological-age gaps, with optimal varying by organ and sex (6.4–7.8 h) 16. Preprint; provides multi-organ resolution beyond the single-clock framing typical of prior studies.

Implication for the U-shape: The pooled epidemiological U-shape (optimum 7 h) replicates across multiple modern cohorts, multi-organ clocks, and MR designs. The new MR work strengthens the causal interpretation for the short-sleep arm of the curve while attenuating it for the long-sleep arm. Mediation through DunedinPACE is now quantified at ~6% of the sleep → mortality effect.

Cross-organism extrapolation

Most foundational sleep-linked glymphatic biology is from rodent models. Humans share AQP4-expressing astrocytes and perivascular anatomy and show sleep-coupled CSF/hemodynamic oscillations, but direct quantification of net endogenous-solute clearance remains technically limited.

DimensionStatusNotes
Pathway conserved in humans?yesGlymphatic system anatomically confirmed in humans; AQP4 present
Phenotype conserved in humans?yesSleep deprivation → inflammation, cognitive impairment well-replicated in humans
Replicated in humans?yes (observational); in-progress (intervention RCTs)Epidemiological signal strong; mechanistic RCTs ongoing

Limitations and open gaps

  • Causal vs correlational sleep-mortality relationship: Long sleep is likely partly explained by reverse causality (illness → more sleep). Short sleep’s causal contribution to mortality vs confounding by lifestyle (shift work, stress) is not fully resolved. gap/contradictory-evidence
  • Optimal sleep duration varies by individual: U-shaped curve pooled across populations; individual optima may differ by genetics (e.g., BHLHE41 “short-sleeper” variants), age, and health status. gap/dose-response-unclear
  • Glymphatic clearance in humans: The Xie result was a young-mouse change from approximately 14% to 23% extracellular-space fraction (approximately 66% relative), not a 60-percentage-point human change. Net sleep-dependent clearance is contested even in mice, and human measurements remain indirect or invasive.34 gap/contradictory-evidence gap/needs-human-replication
  • DTI-ALPS is not direct clearance measurement: A lower index can reflect white-matter geometry, crossing fibers, tract injury, ROI placement, or other local diffusion changes; require an orthogonal fluid-clearance measure before interpreting an ALPS association as impaired glymphatic clearance.8
  • OSA + aging-clock causation: Whether treating OSA with CPAP reverses epigenetic-age acceleration is unresolved. gap/needs-replication
  • Sleep as intervention vs marker: Poor sleep may be downstream of other aging processes (neurodegeneration, chronic pain, nocturia) as much as upstream. Disentangling the direction is essential for intervention design.
  • No head-to-head RCT: CBT-i vs light therapy vs chronotype-alignment vs pharmacological sleep aid on aging-biology endpoints. gap/needs-replication

Cross-references

Footnotes

Footnotes

  1. doi:10.1016/j.neuron.2017.02.004 · Mander BA, Winer JR, Walker MP · Neuron 2017 · review · model: human aging cohort studies + translational review · comprehensive review of SWS/SWA decline with aging: ~75–80% reduction in frontal SWA amplitude in older vs young adults (Figure 1B); spindle density and amplitude decline with age, largest in frontal regions; prefrontal cortical gray matter atrophy predicts severity of SWA impairment; slow-oscillation/spindle uncoupling with age predicts overnight hippocampal-neocortical memory consolidation failure; reviews “do older adults need less sleep?” debate — concludes impaired sleep-generating capacity rather than reduced need ↩ ↩2 ↩3 ↩4

  2. doi:10.1126/scitranslmed.3003748 · Iliff JJ, Wang M, Liao Y, et al. (Nedergaard lab) · Science Translational Medicine 2012;4(147):147ra111 · in-vivo (mouse) · model: adult C57BL/6 mice (8–12 weeks), Aqp4-null + NG2-DsRed + Tie2-GFP transgenics; 2-photon in-vivo microscopy + radiolabeled tracer clearance · discovered paravascular CSF-ISF bulk-flow exchange pathway (named “glymphatic system”) dependent on AQP4 water channels on astrocyte endfeet; CSF enters via periarterial space, exits via perivenous drainage; Aqp4 knockout reduced interstitial solute ([³H]mannitol) clearance by ~70%; demonstrated Aβ clearance via this pathway, reduced ~55% in Aqp4-null mice · verified against full PDF ↩

  3. xie-2013-sleep-glymphatic-clearance · n=77 in the terminal clearance cohort (25 awake, 29 naturally sleeping, 23 anesthetized; 3–6/time point) · in-vivo · model: male C57BL/6 mice, 10–12 weeks · 14.1% awake versus 23.4% sleeping extracellular-space fraction; approximately twofold faster clearance of injected amyloid-β1–40 during sleep; tortuosity null · doi:10.1126/science.1241224 ↩ ↩2

  4. doi:10.1038/s41593-024-01638-y · in-vivo · model: male mice · intraparenchymal small-dye photometry and histology reported lower net dye disappearance during sleep and three anesthetic states; source-data correction doi:10.1038/s41593-024-01698-0 ↩ ↩2

  5. doi:10.1038/s41593-025-01897-3 · 2025 Matters Arising challenging the Miao clearance definition, model, sampling, and statistics · authors’ reply: doi:10.1038/s41593-025-01898-2 ↩

  6. doi:10.1016/j.cell.2024.11.027 · in-vivo multimodal recording and perturbation · model: freely sleeping mice · linked norepinephrine oscillations, slow vasomotion, CSF influx, and clearance; zolpidem suppressed the measured dynamics ↩

  7. thapaliya-2026-me-cfs-dti-alps · n=58 analyzed (31 ME/CFS, 27 healthy controls; 61 enrolled) · cross-sectional · model: adults with physician-diagnosed ME/CFS vs healthy controls · bilateral DTI-ALPS 1.44 ± 0.086 vs 1.51 ± 0.11 (raw p=0.014; p-FDR=0.028); symptom correlations used single self-report items; WHODAS cognition and fatigue associations null · doi:10.3389/fnins.2026.1875420 · PMID 42403482 ↩

  8. dti-alps · Mossige I et al. · n=56 · prospective-observational secondary analysis · model: humans undergoing work-up for cerebrospinal-fluid disorders · DTI-ALPS had limited correspondence with intrathecal gadobutrol-enhanced MRI and no 48-hour tracer-dynamics association; authors favored local white-matter integrity over brain-wide clearance as the index interpretation · doi:10.1148/radiol.252070 · PMID 41631990 ↩ ↩2

  9. doi:10.5665/sleep.4398 · Carroll JE, Carrillo C, Olmstead R, Witarama T, et al. (Irwin lab) · Sleep 2015;38(2):205–211 · in-vivo (human) · experimental crossover (adaptation → baseline → partial sleep deprivation [03:00–07:00 window] → recovery) · n=70 total: 49 older adults (ages 60–84) + 21 younger adults (ages 25–39) · partial sleep deprivation vs baseline sleep; blood collected each morning for monocyte TLR4-stimulated IL-6/TNF-α · age × sleep-loss interaction significant (p<0.05): younger adults showed increased TLR4-stimulated inflammation after sleep deprivation; older adults showed REDUCED baseline TLR4-stimulated inflammation, not further activated by sleep loss · note: PDF download failed (bronze OA, URL mismatch); verified via publisher abstract only — quantitative values approximate ↩ ↩2

  10. doi:10.1161/JAHA.117.005947 · Yin J, Jin X, Shan Z, et al. · Journal of the American Heart Association 2017 · meta-analysis · 67 articles (141 independent reports); n=3,582,016 total participants; 241,107 all-cause deaths · U-shaped dose-response: lowest all-cause mortality at ~7 hr/day; short sleep (<7 hr): pooled RR 1.06 (95% CI 1.04–1.07) per 1-hr reduction; long sleep (>7 hr): pooled RR 1.13 (95% CI 1.11–1.15) per 1-hr increment; cardiovascular events similar U-shaped pattern · nonlinear dose-response analysis; reference category 7 hr ↩

  11. doi:10.1093/sleep/33.5.585 · Cappuccio FP, D’Elia L, Strazzullo P, Miller MA · Sleep 2010 · meta-analysis · n=1,382,999 across 16 studies (27 independent cohort samples); 112,566 deaths · short sleep (≤5–7 hr depending on study) RR 1.12 (95% CI 1.06–1.18); long sleep (>8 hr in most studies) RR 1.30 (95% CI 1.22–1.38) for all-cause mortality · results expressed as RR (relative risk), not OR ↩

  12. Waziry R et al. 2023 — doi:10.1038/s43587-022-00357-y — CALERIE II DunedinPACE result: this was a caloric-restriction trial, not a sleep intervention. Cited here as methodological comparator for DunedinPACE sensitivity. See caloric-restriction § Biomarker effects for full detail. ↩

  13. doi:10.1186/s13148-025-01898-w · Diao T, Liu K, Zhou L, Wang Q, Lyu J, Zhu Z, Chen F, Qin W, Yang H, Wang C, Zhang X, Wu T · Clin Epigenetics 2025;17(1):87 · cross-sectional + prospective cohort (Dongfeng-Tongji 2013–2018) · n=3,566 (mean age 65.5; 426 deaths over 5.4 yr) · sleep score (0–4: bedtime, duration, quality, midday napping) → PhenoAgeAccel β=−0.208, GrimAgeAccel β=−0.107, DunedinPACE β=−0.008, DNAm MS β=−0.019 (all p<0.05); stronger in older adults (interaction p=0.027 DunedinPACE); slower DunedinPACE mediated 6.2% (95% CI 0.8–11.5%) of sleep-score → all-cause-mortality association · PMID 40442824 · PMC12123996 ↩

  14. doi:10.18632/aging.206306 · Wu X, Zhao X, Ge A, Han Z, Hou C, Hao Y, Xiao J, Fan M, Burgess S, Li J, Jiang X · Aging (Albany NY) 2025;17(8):2126–2151 · n=442,664 (UK Biobank) · multivariable linear regression + restricted cubic splines + Mendelian randomization · self-reported sleep duration → PhenoAgeAccel, BioAgeAccel, leukocyte telomere length · observational U-shape with optimum ~7 h/d; MR analyses corroborate deleterious effect of insufficient (short) sleep but not excessive (long) sleep, suggesting long-sleep effect partly reflects reverse causation; cell-type enrichment links short sleep to BioAgeAccel/LTL via muscle and immune-function pathways · PMID 40856643 · PMC12422793 ↩

  15. doi:10.1186/s13148-026-02068-2 · Zhang Z, Wang X, Qiu H, Luo Q, Jiang C, Yang G · Clin Epigenetics 2026;18(1):43 · 2-sample MR (UVMR + MVMR) using GWAS instruments for chronotype, daytime napping, sleep duration · outcomes: telomere length, facial aging, IEAA/HannumAge/PhenoAge/GrimAge, frailty index, cognitive performance · UVMR napping → telomere β=−0.11 (p=0.002), facial aging β=+0.05 (p=0.036), GrimAge β=+0.96 (p=0.048), frailty β=+0.32 (p<0.001); MVMR napping retained associations with GrimAge (β=+1.08, p=0.046) and frailty (β=+0.29, p<0.001); chronotype (morningness) protected facial aging and cognition after MVMR; sleep duration independently protected against frailty (β=−0.36, p<0.001) · PMID 41821113 · PMC12980927 ↩

  16. doi:10.1101/2025.08.08.25333313 · MULTI study; O’Toole CK, Song Z, Anagnostakis F, et al. (Wen J et al., includes Belsky DW, Ferrucci L, Resnick SM, Walker KA) · medRxiv preprint 2025 Aug 11 · cross-omics, cross-organ analysis of sleep duration vs 23 biological-age clocks across 17 organ/tissue systems and 3 omics types (imaging, proteomics, metabolomics) · U-shaped pattern across 9 brain/body systems and 3 omics; optimal sleep range varied by organ and sex (6.4–7.8 h); short and long sleep both associated with increased systemic disease risk and all-cause mortality; MR did not show strong reverse-causal effects from disease to sleep · interactive portal: labs-laboratory.com/sleepchart · preprint; not yet peer-reviewed · PMID 40832429 · PMC12363694 ↩

Graph View

Backlinks