Time matters as much as duration: the same eight hours scattered randomly across the week predicts shorter life and faster cognitive decline than seven hours on a consistent schedule. The levers — light, meal timing, and routine — are cheap.

Circadian rhythms and sleep timing

The circadian system is the body's master timing network: a 24-hour oscillator anchored in the suprachiasmatic nucleus (SCN) of the hypothalamus, with peripheral clocks in nearly every organ. The system is bidirectionally linked to aging — chronological aging dampens circadian amplitude, and circadian disruption itself drives the hallmarks of aging, including mitochondrial dysfunction, deregulated nutrient sensing, and accelerated epigenetic aging. The largest analysis of human sleep regularity to date shows that when you sleep predicts mortality more strongly than how long, and a 2025 Circulation Research state-of-the-art review concluded that sleep irregularity is itself "a robust risk factor" for cardiometabolic disease independent of sleep duration. The leverage here is high and the interventions are cheap.

The master clock and the molecular machinery

The suprachiasmatic nucleus is a cluster of about 20,000 neurons in the mouse — the species in which they have been counted most carefully, the human nucleus being somewhat larger — firing in synchrony to coordinate cellular clocks running in virtually every tissue of the body.^[1] Without external cues, the human period runs slightly longer than 24 hours, averaging about 24.2 h in the forced-desynchrony experiments that isolated it,^[2] so daily synchronisation with environmental signals — zeitgebers — is necessary to keep the system entrained.^[3]

Inside each cell, the rhythm runs on an interlocked transcriptional-translational feedback loop. Two transcription factors, BMAL1 and CLOCK, form a heterodimer that drives transcription of clock-controlled genes — including the PER (Period) and CRY (Cryptochrome) genes that then act as negative regulators, shutting their own production back down. The cycle takes ~24 hours. Secondary loops involving REV-ERBα and RORα stabilise it. The reach is wider than a single percentage suggests. An atlas of twelve organs sampled across the day found 43% of all protein-coding genes rhythmic in at least one of them — but largely organ-specific, so any single tissue carries far less, with liver among the most strongly rhythmic.^[4] (That atlas is mouse; the human picture is less completely mapped.)

This isn't only an alarm clock for sleep. The same machinery gates DNA repair, autophagy, nutrient sensing, and the immune response — meaning circadian disruption isn't a niche concern about feeling jet-lagged; it's a multi-system aging input.

The main zeitgebers

Light. By far the strongest entrainer. Bright morning light advances the clock; bright evening light delays it.
Meal timing. Entrains peripheral clocks more strongly than the master clock — when ten men ate the same meals five hours later for six days, their blood-glucose rhythm shifted by 5.7 hours while melatonin and cortisol rhythms did not move at all.^[5]
Physical activity. A real entrainer, not a weak one: in the largest human phase-response study, an hour of moderate treadmill exercise advanced the clock in the morning and early afternoon and delayed it in the late evening, by amounts the authors called comparable to bright light of the same duration — in a laboratory protocol on aerobically fit adults, not ordinary life.^[6] For sleep specifically, sessions ending four or more hours before bed showed no association with delayed or shortened sleep at any intensity, across four million nights of wearable data.^[7]
Temperature. Core body temperature has its own rhythm; a cool sleeping environment reinforces the natural evening drop and accelerates sleep onset.
Social rhythms. Work, school, and consistent social interaction stabilise the system via routine.

Sleep regularity: the underrated metric (Strong, observational)

The most consequential recent finding is that how regularly you sleep matters more than how long. A 2024 analysis of 60,977 adults in the UK Biobank with seven days of wrist-accelerometer data measured sleep regularity via the Sleep Regularity Index (SRI) — the probability of being in the same sleep / wake state at any two points 24 hours apart.^[8] Comparing the most-regular fifth of adults to the least-regular:

48% lower all-cause mortality (hazard ratio 0.52)
55% lower cardiometabolic mortality (HR 0.45)
39% lower cancer mortality (HR 0.61)

Those are the figures adjusted only for age, sex, and ethnicity, and they are the ones usually quoted. The paper's fully adjusted model — which also accounts for socioeconomic position, lifestyle, and existing health — cuts each roughly in half: 30% lower all-cause mortality (HR 0.70, 95% confidence interval 0.59–0.83), 38% lower cardiometabolic (HR 0.62), and 24% lower cancer (HR 0.76). Still a substantial signal, and still statistically clear, but the headline number is the lightly adjusted one.

Sleep regularity outperformed sleep duration as a mortality predictor in head-to-head models. A 2025 Circulation Research state-of-the-art review concluded that sleep irregularity is independently associated with cardiometabolic disease and likely a stronger predictor than duration.^[9] The same author's primary cohort supplies the dose: across 86,219 adults, each extra hour of night-to-night variation in sleep duration carried about 19% higher cardiovascular risk, independent of inherited risk.^[10] A systematic review of 59 studies puts the mortality signal at 20% to 88% higher in the least-regular sleepers across the five cohorts with the lowest risk of bias — a range across studies rather than a pooled figure, because the metrics were too varied to combine.^[11] All of that is observational. The first randomised attempt to reduce irregularity — 16 adults, twelve nights of enforced regular timing — moved resting heart rate and heart-rate variability in the expected direction but left blood pressure unchanged.^[12] That is the honest state of the evidence: a consistent signal at moderate certainty, with the causal test barely begun. The signal extends across outcomes:

Cardiovascular events: in 72,269 adults over ~8 years, irregular sleepers had a 26% higher rate of major adverse cardiovascular events (HR 1.26). Crucially, adequate sleep duration did not offset the risk of irregularity — you cannot make up for an erratic schedule by logging enough total hours.^[13]
Dementia: a UK Biobank Neurology analysis found the most-irregular sleepers had a 53% higher incident-dementia risk (HR 1.53), with lower gray-matter and hippocampal volume at the regularity extremes.^[14]
Depression and anxiety: regular sleepers had 38% lower incident depression (HR 0.62) and 33% lower incident anxiety (HR 0.67) versus irregular sleepers.^[15]

Onset timing matters alongside regularity. A UK Biobank accelerometer study (N=88,026) found a U-shaped relationship: versus falling asleep at 10:00–10:59 pm, sleep onset at midnight or later carried ~25% higher cardiovascular disease (CVD) risk, before 10:00 pm ~24% higher, and 11:00–11:59 pm ~12% — a defensible sleep-onset window of roughly 10 to 11:30 pm, independent of duration — a range, not a single hour.^[16] Two things keep it broad. Within that cohort the association was stronger in women, and in men only falling asleep before 10 pm reached significance. And a 2025 cohort of 18,129 Chinese adults followed a median 12.4 years reproduced the U-shape on mortality — early sleepers about 18% higher risk, late sleepers about 13% — but placed the lowest-risk bedtime at about 11 pm, and there the effect was stronger in men.^[17] The direction is robust; the hour is not.

The practical implication: Going to bed and waking up at consistent times — including weekends — is at least as important as how long you sleep. "Catching up" on weekends partially restores subjective alertness but does not reverse the weekday metabolic cost. Sleep onset timing irregularity alone has been independently linked to incident hypertension.^[18]

Light: the dominant lever

The total ratio of day-to-night light exposure is the meaningful variable. Maximising daytime brightness and minimising night-time light compounds the effect.

Morning light

Goal: 10–30 minutes of outdoor daylight within an hour of waking.
Outdoor light at midday: 50,000–100,000 lux. Brightest indoor lighting: 500–1,000 lux. The order-of-magnitude gap is the point.
Overcast days still deliver outdoor light roughly 10× brighter than well-lit indoor environments.
Effect: advances the circadian phase and suppresses residual melatonin. The mood benefit is real but demonstrated mainly in people who are depressed rather than in well adults, and the cortisol-awakening claim is the weakest of the four — treat both as plausible rather than established.^[19]

Evening light

Goal: dim, warm, downward-directed light in the two hours before bed.
Evening light suppresses melatonin and delays sleep onset, and the threshold is lower than most people assume. Over a five-hour evening exposure, the average person in a 55-adult dose-response study needed only about 25 lux to lose half their melatonin — dimmer than most living rooms — and the spread between individuals was more than fiftyfold, from about 6 lux at the most sensitive to about 350 at the least. Participants were aged 18–30.^[20]
Practical fixes: warm-temperature LED bulbs (~2700 K or lower), dimmer switches, and avoidance of overhead lighting in the last evening hours. Blue-blocking glasses are the popular answer and the evidence does not support them: a Cochrane review of blue-light-filtering lenses found six randomised trials in 148 people pointing in inconsistent directions, at very low certainty,^[21] and a 2025 meta-analysis measuring sleep with actigraphy rather than questionnaires found no significant effect on any outcome, including the sleep onset the glasses are usually credited with — though on only three crossover trials and 49 people, so treat that as a weak null rather than a demonstration that they do nothing.^[22] Dimming the room is simpler, better evidenced, and free.
Screens at maximum brightness are comparable to overhead room lighting in melatonin-suppressing capacity, but their total share of evening light exposure depends on screen brightness and ambient lighting.

Defensible thresholds (Strong)

A 2022 expert consensus of circadian-lighting researchers set targets in melanopic equivalent daylight illuminance (mEDI) — light weighted to the melanopsin photoreceptors that actually drive the clock, measured vertically at the eye. This is the physiologically correct metric and is more meaningful than raw lux: daytime ≥250 lx mEDI; the three hours before bed <10 lx mEDI; during sleep <1 lx mEDI.^[23] Real-world homes fall well short — modelling of domestic lighting predicted that ~48% of homes would cause at least 50% melatonin suppression.^[24]

Personal light at night and hard outcomes (Strong, observational)

The most direct evidence comes from UK Biobank participants who wore wrist light sensors for a week (~13 million hours of data). Brighter nights raise mortality, and brighter days lower it. In ~89,000 adults followed ~8 years, the brightest-night group (90–100th percentile) had 21–34% higher all-cause mortality than the darkest, while the brightest-day group had lower mortality (HR as low as 0.66); cardiometabolic mortality showed the strongest associations, and suppressed circadian amplitude independently predicted death.^[25] In a parallel analysis (N=84,790, 7.9 y), the brightest-night group had a fully adjusted 53% higher risk of incident type 2 diabetes (HR 1.53, 95% CI 1.32–1.77 — the confidence interval is the range where the true effect most plausibly lies; because it stays above 1.0 here, the increase is unlikely to be a chance finding), dose-dependent across light bands and independent of genetic risk — comparable to the risk conferred by a family history of diabetes.^[26] The same cohort has since been followed for cardiovascular disease specifically, and the pattern holds across every outcome measured: comparing the brightest nights with the darkest, heart-failure risk was about half again as high and heart attack about 50% higher, with coronary disease, atrial fibrillation and stroke all raised by a quarter to a third — in 88,905 adults averaging 62 years old, after adjustment for physical activity, smoking, diet, sleep duration and inherited risk.^[27]

These are observational cohorts from a single accelerometry subsample, so residual confounding and reverse causation remain possible; outdoor (satellite) night-light, by contrast, lost its mortality association after adjustment for air pollution and noise — personal bedroom light is not the same as neighbourhood light. The practical message is robust regardless: minimise night light, maximise day light, keep light–dark patterns regular.

Chrononutrition: when you eat matters (Moderate)

A growing body of evidence shows that meal timing has a measurable metabolic effect independent of calorie content.^[28] Pooled analyses do favour early eating windows — with the last meal closing by 17:00–19:00 — over late ones, but by less than that framing suggests. A network meta-analysis of 41 randomised trials in 2,287 people found early windows beat late ones on body weight by about 1.2 kg and on fasting insulin, both at high certainty.^[29] And the advantage may come from eating less rather than from the clock: a randomised crossover asked 31 women with overweight to keep their usual food and only move an eight-hour window, and neither the early nor the late window improved insulin sensitivity or any other cardiometabolic marker — though intake was not truly held constant, with the early window running about 167 kcal a day lower.^[30]

Mechanisms:

The same meal raises blood sugar more in the evening. Pooling eight randomised crossover trials in 116 healthy adults, the post-meal glucose rise was substantially and consistently larger in the evening than the morning, at moderate certainty — reported as a standardised difference rather than a percentage, so a figure like “50% larger” is not supportable.^[31] Post-meal insulin did not differ by time of day, which means the evening pancreas puts out the same insulin against a bigger glucose load — the same insulin buying less glucose disposal.
The peripheral liver clock anticipates the feeding window; eating outside it produces metabolic friction with hepatic glucose-handling and lipid-synthesis pathways.
The "second-meal effect" is that eating at all improves how you handle the next meal — the same person's post-lunch glucose rise is markedly smaller on a day they ate breakfast than on a day they skipped it, at identical insulin levels, an effect shown even in patients with type 2 diabetes.^[32] That is an argument for eating breakfast rather than for eating it at any particular hour, and it is one reason skipping the first meal fares badly in cohort data.
Late eating wastes fewer calories as heat. Given identical meals at 08:00 and 20:00, the energy burned digesting them in the first two hours — diet-induced thermogenesis — was 44% lower in the evening; a crossover study in 13 adults that separated clock time from body-clock time attributed almost all of that to the clock itself, at 50% lower in the biological evening.^[33] The same food is metabolically more expensive in the morning and more sparing at night — consistent with the isocaloric trial described below, which found late eating lowered waking energy expenditure. Eating close to bedtime also tends to degrade sleep quality, though that evidence is weaker.

A tightly controlled isocaloric crossover trial isolated the timing effect from calories: eating the same meals four hours later decreased waketime energy expenditure (~59 fewer kcal burned per waking day), lowered 24-hour core temperature, increased hunger, shifted the ghrelin:leptin ratio, and pushed adipose gene expression toward lipid storage — explaining why late eating harms even when intake is matched.^[34] In a large cohort, a later first meal was associated with ~6% higher cardiovascular risk per hour of delay, arguing specifically against breakfast skipping.^[35]

Practical implication: larger breakfast / lunch, lighter and earlier dinner. The familiar “finish eating 2–3 hours before bed” comes from gastroenterology rather than chronobiology — it is a conditional recommendation in the reflux guidelines, on low-quality evidence.^[36] It is sensible, but it is not a circadian finding. See Fasting and time-restricted eating and Glycemic index for the full mechanistic picture.

Peripheral clocks and the consequences of internal desynchrony (Moderate, mechanistic)

The SCN is the conductor; the orchestra is the cellular clocks running in virtually every tissue of the body. Each one runs on the same BMAL1/CLOCK/PER/CRY machinery, but with a tissue-specific transcriptome under its control. Liver clocks gate gluconeogenesis, lipogenesis, and cholesterol metabolism. Cardiac clocks gate the rhythmic preference for fatty-acid versus glucose oxidation. Immune clocks gate inflammatory tone and lymphocyte trafficking. Whether they gate vaccine response is less settled than it is often said to be: one cluster-randomised trial in adults over 65 found a morning advantage in antibody response for two of three influenza strains, but practices were allocated unevenly and the third strain showed nothing.^[37]

When central and peripheral clocks fall out of phase — internal desynchrony — the consequences spread across systems. Liver-specific clock disruption produces hepatic steatosis and insulin resistance. Cardiac clock disruption predisposes to arrhythmia and contributes to heart failure with preserved ejection fraction. The obvious inference — that you should time your medication to these clocks — has been tested at scale and failed; see blood pressure for the two large randomised trials that settled it.

The BMAL1–mTOR axis

The same BMAL1 that runs the cellular clock also acts as a negative regulator of mTOR complex 1 (mTORC1) — the nutrient-sensing kinase whose chronic activation drives aging (see Protein). In BMAL1-deficient mice, mTORC1 activity rises uncontrollably and the animals develop a premature-aging phenotype (sarcopenia, cataracts, loss of subcutaneous fat). The mTOR inhibitor rapamycin extends those mice's median lifespan by about half — from roughly 7.8 to 11.5 months — partially rescuing the phenotype.^[38] Read that carefully: these are animals engineered to age prematurely and die early, so a 50% extension restores part of a truncated lifespan rather than extending a normal one. It is good evidence that circadian control of nutrient sensing matters mechanistically, and no evidence at all about healthy animals, let alone people.

The SIRT1–NAD⁺ feedback loop

CLOCK and BMAL1 also drive rhythmic expression of NAMPT, the rate-limiting enzyme in NAD⁺ biosynthesis — which in turn drives rhythmic activity of the SIRT1 deacetylase. SIRT1 feeds back by deacetylating PER2 and BMAL1, modulating their stability. As we age, NAD⁺ levels fall, this feedback dampens, the circadian amplitude flattens, and the system loses its capacity to coordinate metabolism with the light–dark cycle. The clock and the longevity-relevant nutrient sensors are not separate stories — they are mechanistically interlocked.

Feeding timing as a longevity lever

The strongest mechanistic case for circadian-aligned eating comes from a landmark lifespan study: in male mice, 30% calorie restriction alone extended lifespan by 10%, but confining the same restricted intake to a daily fasting interval aligned with the active phase extended it by 35% — independent of body weight.^[39] When the calories are eaten, not just how many, is the lever. (Mouse data; not yet demonstrated in humans.)

Hormonal rhythms and how they change with age (Moderate)

Several hormones run on tightly entrained circadian rhythms whose timing and robustness change with age, and in some cases their amplitude:

Cortisol peaks in the early morning (the "cortisol awakening response") and declines through the day, and it is the shape of that decline that carries the prognostic signal rather than the peak. Among 4,047 British civil servants followed six years, a flatter daily slope predicted death — about 30% higher all-cause mortality per standard deviation of flattening, against a background of roughly one death in thirty — with a larger increase in cardiovascular death that rested on only 32 such deaths. Morning cortisol and the awakening response predicted nothing.^[40]
Melatonin is secreted by the pineal gland during the dark phase, and its decline with age is smaller than folklore holds. The steep fall happens from childhood into early adulthood; among healthy older adults it may not happen at all. When 34 adults aged 65–81 were measured under controlled constant conditions against 98 young men, 24-hour mean melatonin was essentially identical — 70 versus 73 pmol/L, with the same nocturnal peak height and duration.^[41] What ages is usually the rhythm's timing and robustness rather than its amplitude. Melatonin is also widely described as a mitochondrial antioxidant. A critical commentary in the field's own journal accepts that it scavenges free radicals in cell-free chemistry but argues that direct scavenging inside living tissue is neither theoretically plausible nor demonstrated — proposing instead that what melatonin actually does is induce the body's own antioxidant enzymes such as glutathione peroxidase and superoxide dismutase, which is the part of the claim that survives.^[42]
Thyroid-stimulating hormone (TSH) is released in a 24-hour rhythm that peaks overnight. A frequent-sampling study in eight older men found both their levels lower and their overnight rise about an hour earlier than young men's. Don't read that as older people having low TSH — at population scale the distribution shifts upward with age, which is why age-specific reference ranges matter when reading a result. See Thyroid management.
Growth hormone is secreted in pulses tied to slow-wave sleep, and the two decline together. Across 149 healthy men aged 16 to 83, deep sleep averaged about 19% of the night in the 16–25 group and about 3% in the 36–50 group, with growth-hormone secretion lower in step — a comparison across ages rather than within people.^[43]
Testosterone peaks in the early morning, especially in younger men. The morning peak attenuates with age; see Testosterone therapy.

The cumulative picture: an aging body is one whose internal hormonal rhythms are losing phase precision. Maintaining circadian inputs — light, meal timing, sleep regularity — is one of the few non-pharmacological levers that touches multiple hormones at once.

Chronotype (morning lark vs night owl) is partly genetic, but not in the way the popular account suggests: a genome-wide study of 697,828 people found 351 separate spots in the genome each contributing a tiny amount, and carrying the most morningness variants rather than the fewest shifts sleep timing by only about 25 minutes.^[44] The rest is lifelong light-exposure history. Being an evening type carries a real if modest cost: in 433,268 UK Biobank adults, definite evening types had about 10% higher all-cause mortality than definite morning types — an extreme-versus-extreme comparison on a single self-reported question.^[45] Social jetlag — the difference between sleep timing on workdays and free days — independently predicts cardiovascular disease: two hours or more of it carried about 30% higher risk in 51,562 adults, holding after adjustment for sleep duration and inherited cardiac risk, though that comes from a single very recent cohort.^[46] Its link to depression is real but small enough to be clinically marginal, and there is no evidence at all that it shortens life expectancy.^[47]

A modest forced shift toward an earlier schedule is usually feasible:

Anchor wake time first; bedtime shifts naturally to follow.
Front-load morning light exposure.
Size the caffeine cut-off to the dose, not the clock. A single cup (about 100 mg) had no measurable effect on sleep even four hours before bed in a crossover trial of 23 young men, while 400 mg in one hit disturbed sleep from twelve hours out and worsened the closer it came.^[48] A meta-analysis of 24 studies put the margin needed to protect total sleep time at roughly nine hours for a normal coffee and thirteen for a pre-workout serving.^[49] A heavy load wants the 10–12 hour buffer; one late-afternoon cup is defensible for most people.
Avoid late-night eating, which delays peripheral clocks.

When it stops being a preference. Some people cannot shift: a night owl who reliably cannot fall asleep before 03:00 has delayed sleep-wake phase disorder, and an older adult falling asleep at 19:00 and waking at 03:00 has its mirror image, advanced sleep-wake phase disorder. Both are diagnosable conditions rather than habits, both have specific treatments — timed light, timed low-dose melatonin, structured schedule shifting — set out in the American Academy of Sleep Medicine's 2015 clinical guideline, still the governing document, and both warrant a referral rather than more sleep hygiene.^[50]

Circadian disruption: shift work and jet lag (Strong, observational)

Night shift work is classified by the World Health Organization's International Agency for Research on Cancer (IARC) as probably carcinogenic to humans (Group 2A).^[51] That verdict rests on three legs, and it is worth knowing which is which: limited human evidence — meaning credible but not immune to confounding — for cancers of the breast, prostate, colon and rectum; sufficient evidence in animals; and strong mechanistic evidence, also in animals. The cardiovascular risk is better quantified: pooling 21 studies and 173,010 people, shift workers had about 17% more cardiovascular events, with coronary events raised by about a quarter and cardiovascular death by about a fifth, and risk climbing roughly 7% for every additional five years of shift work — but only past the first five, below which the pooled data show no excess.^[52] Type 2 diabetes, depression and metabolic syndrome are also elevated.

Jet lag is functionally the same physiology over a shorter timescale. Recovery is roughly one day per time zone crossed. Eastward travel (phase advance) is harder than westward (phase delay) because the human circadian period naturally runs slightly longer than 24 hours.

Mitigation strategies with moderate evidence:

Pre-travel and on-arrival light exposure timed to the destination's morning.
Low-dose immediate-release melatonin (0.5–5 mg) at the destination's bedtime for 3–5 days. This is the best-supported jet-lag intervention there is — the Cochrane review puts the number needed to treat at 2 for travellers crossing five or more time zones, and finds doses above 5 mg no better than low ones. Formulation matters more than brand: the same review found slow-release 2 mg relatively ineffective, so prolonged-release products are the wrong choice here despite their more reliable content.^[53] US over-the-counter content is genuinely unreliable — one analysis found actual melatonin ranging from 83% below to 478% above label, with lot-to-lot variation within a single product reaching 465%.^[54] The problem has not improved: a US Food and Drug Administration survey of 110 melatonin products marketed for children found melatonin in 108 of them at anywhere from 0% to 667% of the declared amount.^[55]
Strategic caffeine in the destination's morning, avoided in the evening.
Outdoor moderate-intensity exercise on arrival, especially in morning light.

Clock changes are the version of this that happens to everyone involuntarily, twice a year. Pooling 12 studies across 10 countries, heart attacks rose about 4% in the days after the spring transition, when an hour of sleep is lost; the autumn transition showed no clear effect.^[56] Sleep-medicine bodies now argue that keeping clock time closer to solar time all year would be healthier, precisely because it stops imposing a twice-yearly phase shift on an entire population.^[57]

Practical guidance

Anchor a consistent wake time, weekends included. The Windred et al. mortality data is built on this single behavioural variable. Same wake time → same bedtime drift.
Get morning sunlight within an hour of waking — 10 minutes is enough on a sunny day; longer on overcast days.
Dim and warm the evening. Reduce overhead lighting and screens in the last two hours before bed.
Keep the bedroom cool — but don't chase a specific number. Cool enough that core temperature can fall is the principle; the often-quoted 18–20 °C (65–68 °F) is convention rather than evidence. When older adults were monitored at home with environmental and wearable sensors, sleep was most efficient and restful at 20–25 °C (68–77 °F), falling 5–10% as bedrooms warmed from 25 to 30 °C.^[58] A warm bath or shower (40–42.5 °C) one to two hours before bed exploits the same mechanism in reverse — distal-skin vasodilation accelerates the core-temperature decline that gates sleep, advancing sleep onset by ~10 minutes on average.^[59]
Eat earlier. Larger breakfast and lunch, lighter and earlier dinner. Finishing 2–3 hours before bed is a reflux recommendation rather than a circadian one, but it is sensible anyway.
Don't rely on the weekend to undo the weekday. In a randomised in-laboratory study, 14 healthy young adults on a repeating pattern of weekday restriction plus ad-libitum weekend recovery ended up with insulin sensitivity 9% to 27% lower depending on whether whole-body, liver or muscle was measured, along with delayed circadian phase, more after-dinner eating and weight gain — the weekend prevented none of it.^[60] Sleeping in is not itself harmful, though: across 73,513 adults followed eight years, two or more hours of weekend catch-up sleep was associated with neither higher nor lower mortality or cardiovascular disease.^[61]
For chronic shift workers, the harm is real and partial mitigation is possible — strict bedroom blackout during day sleep, melatonin-timed adaptation, and minimizing rotating-shift exposure where possible.

What's overrated

"Catching up on weekends." Partial subjective recovery; the metabolic deficits don't reverse.
Blue-light-blocking glasses worn all evening as a sole intervention. No effect on sleep onset when it is measured objectively rather than by questionnaire; dimming the room matters more.
High-dose evening melatonin (3–10 mg) — though “the right dose” depends on what you want. A dose-response meta-analysis of 26 randomised trials found the effect on falling asleep and total sleep time peaking around 4 mg, and given about three hours before bed rather than the usual half-hour.^[62] The strongest predictor in that analysis was whether the person had insomnia at all, so read 4 mg as an optimum for treating insomnia rather than a target for a well sleeper. And 0.3 mg is what restores physiological blood levels, whereas 3 mg spills melatonin into the following day — so a small dose is the right choice for shifting the clock, and the larger doses buy sleep-onset minutes at the cost of next-day effects. Neither justifies 10 mg. A preliminary 2025 analysis also linked long-term daily melatonin use for insomnia to higher heart-failure and mortality rates (HR ~1.9–2.1), which tempers any "harmless" framing — though it is an unpeer-reviewed conference abstract in a chronic-insomnia population (where the underlying poor sleep itself raises cardiovascular risk) with no dose data.^[63] The occasional low-dose use described above remains well supported.
Light therapy for healthy adults at temperate latitudes. Genuine outdoor morning light is brighter than most light boxes and free. Light boxes earn their place in clinical populations, and the list is longer than seasonal depression: a meta-analysis of 11 randomised trials found bright light roughly doubled remission and response rates in non-seasonal depression,^[64] and it also improves sleep and agitation in people living with dementia.^[65] The bullet stands for well adults; it does not generalise past them.
"Anti-blue-light" filters on screens during the day. They reduce the bright daytime light that the SCN actually needs to entrain to. Use them in the evening or not at all.

Circadian rhythms and sleep timing

Circadian rhythms and sleep timing

The master clock and the molecular machinery

The main zeitgebers

Sleep regularity: the underrated metric (Strong, observational)

Light: the dominant lever

Morning light

Evening light

Defensible thresholds (Strong)

Personal light at night and hard outcomes (Strong, observational)

Chrononutrition: when you eat matters (Moderate)

Peripheral clocks and the consequences of internal desynchrony (Moderate, mechanistic)

The BMAL1–mTOR axis

The SIRT1–NAD⁺ feedback loop

Feeding timing as a longevity lever

Hormonal rhythms and how they change with age (Moderate)

Chronotypes and social jetlag (Moderate, observational)

Circadian disruption: shift work and jet lag (Strong, observational)

Practical guidance

What's overrated

Further reading

More in Sleep