Light at night and metabolic health: one dose-response story from bedroom lamp to night shift

Summary

Light in your bedroom at night predicts type 2 diabetes dose-dependently in the largest objective-sensor study ever run — the brightest tenth of nights carries a hazard above the average night-shift job — and a single experimental night at 100 lux measurably worsens next-morning insulin resistance through a pathway that does not require you to wake up or even suppress melatonin; the fix costs nothing, and the honest ceiling is Moderate because the entire disease arc so far comes from one cohort analysed by one research consortium.

Why Moderate

This entry sits at Moderate.

Why not the tier above (Strong). Every disease outcome in the arc — diabetes, mortality, cardiovascular disease, depression — derives from one cohort analysed by one consortium; however large and well-measured, that is one draw from the world, and this platform does not grade single-cohort observational arcs Strong. The exposure window is one week standing in for years. The causal bridge is twenty people, one night, surrogate endpoints, one lab. Reverse causation has an untestable residue. Strong requires independent replication that does not yet exist, and the entry names that explicitly rather than treating within-cohort outcome breadth as if it were replication.

Why not the tier below (Emerging). This is no longer a hypothesis literature. The exposure measurement is objective and prospective at a scale (85,000–89,000 people) the field has never had before; the diabetes association is dose-responsive and independent of polygenic risk; the direction was predicted in advance by circadian biology, with the harmful night half present across all four outcome families and the protective day half on mortality and mood; a randomised human experiment moves the relevant physiology; a rodent model closes the mechanism loop; and the whole arc is coherent with the Strong-tier shift-work literature it gradates into. That triangulation is exactly what separates Moderate from Emerging on this platform.

Per-sub-area split. The sensor cohort arc: Moderate for the prospective diabetes and mortality arms; Moderate-to-Emerging for the cross-sectional depression arm; Emerging for the 2025 cardiovascular extension. The experimental bridge (Mason, Cheung): Emerging — randomised and real, small and surrogate. The rodent mechanism (Fonken): Emerging as mechanism support, inadmissible as human evidence. The shift-work anchor: Strong, owned by shift_work_circadian_health. The gradient note comparing bedroom light to shift work: Moderate as a direction, never to be quoted as a precise cross-study comparison.

Practical takeaway

Everything with evidence behind it in this entry is free. That is stated as fact, not as virtue: the cost-inversion across the whole light literature — the strongly evidenced interventions cost nothing, the expensively marketed ones sit at the weak end — is one of the platform's clearest worked examples, and this entry is its sharpest case.

The bedroom: genuinely dark, not dimly lit.
• Target real darkness — as a working test, you should struggle to see your hand at arm's length. The experimental harm signal appeared at 100 lux (an ordinarily lit room); the cohort gradients run down into much dimmer exposure; the rodent mechanism operated at 5 lux. There is no established human "safe threshold," so the practical answer is: as dark as your life allows.
• Lights off means off, not down. Blackout curtains or blinds if streetlight enters. Cover or remove standby LEDs — individually trivial, collectively a lit room to a dark-adapted eye.
• A sleep mask is the complete fix for a few pounds if the room cannot be darkened (shared rooms, travel, summer latitudes). The light pathway is through the eyes; a mask closes it.
• TVs and phones do not belong on during sleep. Not because of their spectrum — because they are light sources running through your eyelids all night.

The evening: dim and warm beats bright, hours before bed. Drop overhead brightness in the last couple of hours; lamps below eye level; whatever warm/dim settings your existing lights already have. This serves the circadian system generally — the protocol detail is owned by circadian_rhythm_optimization — but it also directly reduces the exact exposure class the evening glucose experiment used.

The day: get bright light, ideally outdoors. Brighter days ran protective on the mortality and mood outcomes in the same analyses that indicted bright nights, and daytime light strengthens the rhythm that night light perturbs. Morning outdoor light is the highest-leverage slot (owned by the circadian entries).

Night waking: if you need to navigate at 3am, use the least light that keeps you safe, placed low, switched on briefly — not the ceiling light. (Safety overrides all of this; see below.)

What NOT to buy off the back of this entry: blue-blocking glasses (the Cochrane review found probably no eyestrain benefit and indeterminate sleep evidence), "circadian" bulb systems, spectrum-engineered night lighting, or any product whose pitch borrows this entry's cohort statistics. The evidence here is about presence and timing of light. Darkness needs no spectrum engineering.

Evidence detail

Why This Entry Exists

For most of the history of light-and-health content, the disease claims lived entirely at the wellness end of the spectrum: junk-light fear, blue-light panic, special glasses, special bulbs. The scientific consensus said light timing matters for sleep and circadian rhythm — and stopped there. That changed between 2022 and 2025, and this entry exists to state precisely what changed, how far it goes, and where it stops.

What changed is measurement. Earlier light-at-night epidemiology relied on satellite images of neighbourhoods or one-line questionnaires — proxies so crude that any association could be argued away. Then a research consortium put wrist-worn light sensors on roughly 85,000 to 89,000 UK Biobank participants for a week each and followed them prospectively for years. The result is a single continuous dose-response story that runs from the bedside lamp to the night shift: brighter nights predict incident type 2 diabetes in a stepwise gradient, predict higher all-cause mortality, predict incident cardiovascular disease, and associate with depression — while brighter days predict the opposite direction on the mortality and mood outcomes (in the cardiovascular analysis, the day-light protective association did not survive full adjustment). Alongside it sits a small experimental literature showing that the acute physiology moves under randomisation: one night of sleeping under 100 lux of room light — eyes closed, asleep — raised next-morning insulin resistance in healthy volunteers.

This entry also exists because both available distortions of that finding are already circulating. One direction inflates it into "light at night causes diabetes" and uses it to sell spectrum products. The other direction dismisses it with the older, correct-at-the-time line that light-at-night disease claims are wellness overclaims. Neither survives contact with the current evidence, and the honest middle — dose-responsive, prospectively measured, objectively sensed, causally bridged, and still resting on one cohort — is what this entry holds.

What bad advice this protects against, in all directions:
• "A bit of light while you sleep is harmless — your eyes are closed and you're unconscious." The best experiment on this question was designed to test exactly that assumption. Participants slept — asleep, eyes closed — under 100 lux of ordinary room light for one night, and their insulin resistance the next morning was measurably worse than the dim-condition group's. Their melatonin was unchanged, and the effect tracked with elevated heart rate and a shift toward sympathetic nervous system dominance during sleep. Closed eyelids attenuate light; they do not make it biologically silent.
• "I fall asleep fine with the TV on, so it isn't affecting me." Falling asleep easily answers a question about sleep onset, not about this pathway. The experimental effect above occurred in people who slept. The mechanism is not "light kept you awake"; it is that light reaching the retina through closed lids kept part of the nervous system in a daytime state while the body was trying to run night metabolism.
• "Only night-shift workers need to worry about light and metabolic disease." The gradient runs continuously, and its top end is not where people assume. In the sensor cohort, the brightest tenth of ordinary bedroom nights carried a type 2 diabetes hazard ratio of 1.53 — higher than the average hazard the shift-work meta-analytic literature assigns to actually working nights (around 1.30). Shift work is the extreme of a shared axis, not a separate category of person.
• "Light at night is proven to cause diabetes — it's the hidden driver of the metabolic epidemic." This is the overclaim in the other direction, and it matters just as much. The disease-outcome evidence is observational; one week of light data stands in for years of exposure; the people who are ill, in pain, or anxious are awake with the lights on more, so reverse causation is live; and every disease-outcome analysis so far comes from the same cohort analysed by the same consortium. The experimental bridge is real but small, short, and measured on surrogate markers, not disease. "Dose-responsive risk marker with a plausible causal bridge" is the defensible sentence; "proven cause" is not.
• "Protect your metabolism at night — get blue-blocking glasses and amber bulbs." The evidence in this entry is about light presence and timing, not spectrum, and darkness is free. The best systematic review of blue-blocking lenses found they probably do nothing for eyestrain and the sleep evidence is indeterminate. Products borrow this entry's pattern evidence to sell a spectrum solution the pattern evidence does not test and does not need.

What this entry OWNS: the light-at-night-to-metabolic-and-disease-risk evidence — the objective-sensor cohort arc (diabetes, mortality, cardiovascular disease, mood), the acute experimental bridge (the 100-lux night; evening blue-enriched light and glucose), the rodent mechanism, the bedroom-to-shift-work gradient note, and the free practical protocol (dark bedroom, dim evening, bright day) as it relates to metabolic risk.

What this entry DEFERS: occupational night-shift detail — rotation patterns, mitigation strategies, the full shift-work disease literature — belongs to shift_work_circadian_health. How darkness serves sleep quality itself belongs to darkness_and_sleep. Glucose regulation as a system, and insulin resistance as a condition, belong to blood_sugar_regulation and insulin_resistance_and_metabolic_dysfunction. Building the daytime half of the light rhythm belongs to circadian_rhythm_optimization. When you eat relative to your clock belongs to meal_timing_consistency_and_circadian_alignment. The blue-blocker product question belongs to blue_light_blocking_evening. The family-wide tier map and the doctrine on light claims in general belong to light_environment_and_health.

Evidence

Read the tiers, not the thesis. The headline claims of this entry sit at Moderate; the experimental bridge under them is Emerging; the rodent mechanism is rodent and stays labelled that way; and the occupational anchor is Strong but owned by another entry. The per-claim labels below are the entry's actual content — the thesis is only their sum.

Sub-area A: The objective-sensor cohort arc (Moderate — large, prospective, dose-responsive, and all from one cohort)

1. [Moderate — human observational, prospective] In 84,790 UK Biobank participants who wore wrist light sensors for one week and were followed prospectively, brighter night-time light exposure predicted incident type 2 diabetes in a stepwise dose-response: hazard ratios of 1.29, 1.39 and 1.53 across increasingly bright night-exposure groups compared with the darkest, with the brightest tenth of nights carrying the 1.53. Critically, the association was independent of polygenic risk for diabetes — light at night predicted incident disease at every level of genetic susceptibility, which removes one of the most obvious confounding stories (that genetically susceptible people also happen to live differently). Adjustment included lifestyle and socioeconomic factors; it cannot include everything. Windred et al. 2024, The Lancet Regional Health – Europe. Cui bono: nothing is sold here — darkness is free, which is a genuine point for credibility. The pull the other way is reputational rather than commercial: the consortium that built this exposure dataset has a career stake in its findings mattering.

2. [Moderate — human observational, prospective] In 88,905 participants of the same cohort, brighter nights predicted higher all-cause mortality — 15% to 34% higher across the brighter-night exposure groups and adjustment models, with the brightest tenth of nights carrying the 21–34% band — while brighter daytime light predicted lower mortality. The two-directional pattern — night light harmful, day light protective — is the signature this arc's mortality and mood analyses share (the cardiovascular arm, below, retained only the night half after full adjustment), and it is what you would predict if circadian disruption, rather than light itself, were the active variable. Windred et al. 2024, PNAS. Cui bono: as above — no product on the harmful side; note that "bright days are protective" is the one half of the finding the lighting industry can and does monetise.

3. [Moderate/Emerging — human observational, cross-sectional] In 86,772 participants of the same cohort, brighter night light was associated with roughly 30% higher odds of major depressive disorder, and brighter day light with roughly 20% lower odds. This analysis is cross-sectional — exposure and outcome measured in the same window — so it is a weaker design than the prospective arms above and is labelled down accordingly. Depression disturbs sleep and puts people under lights at night; the direction of travel cannot be settled from this design. Burns et al. 2023, Nature Mental Health. Cui bono: light-therapy device makers benefit from the day-light half; the night half sells nothing.

4. [Emerging — human observational, prospective; verified against the primary record for this entry] In the same 88,905-participant sensor cohort followed for an average of 9.5 years, the brightest tenth of nights (compared with the darkest half) predicted higher incidence across the cardiovascular disease family: roughly 23–32% higher for coronary artery disease, 42–47% higher for myocardial infarction, 28% for stroke, 45–56% for heart failure, and 28–32% for atrial fibrillation (each range spanning the minimally to fully adjusted models). Notably, the day-light half of the signature did not hold here: brighter daytime light showed protective associations only in minimally adjusted models, and those disappeared after adjustment for lifestyle — in the fully adjusted analysis, day light was not associated with any cardiovascular outcome. This extends the outcome family for night light; it does not add cohort independence, it weakens the universality of the bidirectional story, and it is the newest and least-digested arm of the arc, so it carries the lowest label in this sub-area. Windred, Burns, Rutter et al. 2025, JAMA Network Open. Cui bono: as above.

5. Structural caveat — and it binds every claim in this sub-area. All four analyses come from one cohort (UK Biobank) analysed by one collaborating consortium. That is not a flaw in the work; it is a limit on how much the work can currently prove. One week of sensor data stands in for years of habitual exposure. UK Biobank volunteers are healthier and wealthier than the population. And reverse causation is live in a way adjustment cannot fully kill: people who are ill, in pain, anxious, or caring for others are awake with the lights on more, before any disease is diagnosed. Dose-response gradients and polygenic independence make the confounded reading harder, not impossible. Independent replication in a different cohort with objective sensors is the single thing this literature most needs, and this entry says so rather than rounding up.

Sub-area B: The experimental causal bridge (Emerging — real, randomised, small, short, surrogate)

6. [Emerging — human experimental] Twenty healthy adults were randomised to sleep one night under 100 lux of overhead room light — roughly a lit bedroom, well below overhead-office brightness — or under dim light (<3 lux). The room-light group showed higher insulin resistance the next morning. Two details carry most of the meaning. First, melatonin was unchanged between conditions — the popular "light suppresses melatonin, melatonin protects metabolism" story cannot explain this result. Second, the lit group showed elevated heart rate and a shift toward sympathetic (fight-or-flight-side) nervous activation during sleep, and the metabolic effect tracked that activation: the body ran the night partly in a daytime autonomic state. Participants were asleep with their eyes closed throughout. This is one night, twenty people, one lab, surrogate endpoints — it is a causal bridge, not causal proof of disease. Mason et al. 2022, PNAS. Cui bono: publicly funded academic sleep laboratory; the finding sells nothing except turning your lights off.

7. [Emerging — human experimental] In 19 healthy adults in an inpatient protocol, three hours of blue-enriched bright light in the evening produced a higher glucose peak in response to a meal compared with dim light. Acute, small, surrogate — but a second independent experimental demonstration that ordinary-intensity evening light moves glucose handling in humans within hours. Note the honest limit of the "blue-enriched" label: this study manipulated bright, blue-enriched light versus dim, so it cannot separate intensity from spectrum — do not cite it as proof that blue specifically is the problem. Cheung et al. 2016, PLoS One. Cui bono: none identified on the study; downstream, blue-blocker sellers routinely cite this design class as if it isolated spectrum, which it did not.

Sub-area C: The mechanism layer (rodent — never to be presented as human evidence)

8. [Emerging — rodent] Mice housed under dim light at night — 5 lux, roughly a nightlight — gained significantly more weight than mice in dark nights despite equivalent total food intake and activity, and the driver was when they ate: dim night light shifted food intake into the biological rest phase, and restricting food to the correct phase prevented the weight gain. This is the cleanest available demonstration of the proposed mechanism — light at the wrong time misaligns feeding with the metabolic machinery evolved to expect food at particular clock phases — and it is a mouse finding, full stop. It licenses a mechanism hypothesis in humans; it licenses no human claim by itself. Fonken et al. 2010, PNAS. Cui bono: none; academic circadian biology.

Sub-area D: The occupational anchor and the gradient (Strong for shift work — detail owned elsewhere)

9. [Strong — human observational, meta-analytic; detail defers to shift_work_circadian_health] Shift work — the extreme, decades-studied end of light-and-activity at the wrong biological time — carries a meta-analytic hazard ratio of roughly 1.30 for incident type 2 diabetes, with the metabolic-syndrome literature concurring. This entry cites it for one purpose only: the coherence check. BMC Endocrine Disorders 2024 meta-analysis; Sooriyaarachchi et al. 2022, Obesity Reviews.

10. [Moderate — the gradient note; a synthesis across claims 1 and 9, not a quoted result] Put claims 1 and 9 side by side: ordinary people whose bedrooms sit in the top tenth of night-light exposure carry a diabetes hazard (1.53) that exceeds the average hazard of actually working night shifts (about 1.30). The comparison is between different studies with different designs and should not be pressed for precision — but its direction is the point. The risk axis does not begin at the factory door. It runs continuously from the streetlight through the curtains, through the TV left on, to the rotating night shift, and the exposure most people can actually change is at the cheap end of that line.

Mechanism

The clock does not have an off switch at night — it has a "keep it dark" requirement. The master circadian clock in the hypothalamus takes its primary timing signal from retinal cells (intrinsically photosensitive ganglion cells) that respond to ambient light and report to the brain independently of conscious vision. They report through closed eyelids at reduced but non-zero strength. Light arriving during the biological night is not ignored because you are asleep; it is a daytime signal delivered at the wrong phase.

The best human experiment points at the autonomic nervous system, not melatonin. The popular mechanism story — light suppresses melatonin, and melatonin loss damages metabolism — is not what the strongest experiment found. In the 100-lux night study, melatonin did not differ between conditions, but heart rate rose and heart-rate variability shifted toward sympathetic dominance during sleep, and next-morning insulin resistance tracked that shift. The working model: light at night holds part of the nervous system in a daytime alert state, and a body running sympathetic activation through the night handles glucose worse the next morning. This matters practically, because it means "my melatonin is fine" and "I slept fine" do not answer the exposure question.

Peripheral clocks and mistimed intake complete the picture — in animals. Nearly every metabolic tissue runs its own subordinate clock; insulin sensitivity, and the machinery for processing food, are scheduled. The rodent work shows dim night light shifting when animals eat into the rest phase, producing weight gain that correct-phase feeding prevents. In humans this layer is hypothesis supported by analogy and by the separate meal-timing literature (owned by meal_timing_consistency_and_circadian_alignment) — it is plausible, coherent, and not directly demonstrated as the pathway behind the cohort findings.

Why day-bright and night-dark move together — mostly. In the mortality and mood analyses, bright days predicted better outcomes where bright nights predicted worse; the cardiovascular analysis is the exception, where day-light protection did not survive adjustment for lifestyle. A strong daytime light signal deepens the amplitude of the circadian rhythm — a clock with a strong day signal is harder to perturb at night. This is why the practical protocol below is a rhythm protocol, not just a darkness protocol — while noting honestly that the protective-day half of the evidence is currently thinner than the harmful-night half.

Risks And Contraindications

Darkness is one of the safest interventions the platform describes, and its one real risk is mechanical: falls. For older adults, anyone on sedating medication, and anyone unsteady at night, a pitch-dark unfamiliar path to the bathroom is a genuine fracture risk that outweighs a marginal metabolic exposure. The correct trade is a low-placed, motion-activated, dim nightlight along the floor path — not restored room lighting, and not stoicism in the dark. A sleep mask plus a lit path is a reasonable resolution of the conflict.

Do not let this entry manufacture anxiety about unavoidable or trivial exposure. One lit night, a hotel room, a partner's reading lamp, a summer dawn — the cohort findings describe habitual exposure gradients across years, not single nights. Health anxiety about light will cost more sleep than the light.

Shift workers should not read this entry as "your job is destroying you, quit." The occupational evidence, its honest effect sizes, and the mitigation playbook belong to shift_work_circadian_health. What this entry adds for a shift worker is one useful reframe: the bedroom half of the gradient is under your control even when the work half is not, and dark daytime-sleep environments are the highest-leverage free move available.

This entry is not diabetes management advice. Anyone with diagnosed diabetes or prediabetes adjusts monitoring and medication with their clinician; light hygiene is an environmental layer under blood_sugar_regulation, never a substitute for treatment.

Morning bright-light therapy (for mood or rhythm) is not contradicted by this entry. Timing is the entire point: the same light that harms at 2am is the protective signal at 8am.

Controversy

Nature of the dispute. Nobody seriously disputes the associations exist — the sensor data is objective and the gradients are clean. The dispute is what they mean: is light at night an independent causal contributor to metabolic and cardiovascular disease, or a marker that efficiently tags everything else already going wrong — disrupted sleep, illness, pain, shift work, social disadvantage, late eating? A second, quieter dispute is how much weight a single cohort can carry, however large.

Position A: light at night is a causal disease-risk factor. Best evidence: the dose-response gradient across exposure levels, present after adjustment; independence from polygenic diabetes risk; prospective design with objective sensors (killing recall bias); the bidirectional signature (night light harmful, day light protective) predicted in advance by circadian biology and present on the mortality and mood outcomes; a randomised human experiment in which one lit night moved insulin resistance through a measured sympathetic pathway; and a rodent model demonstrating the mechanism end-to-end. That is the triangulation pattern — epidemiology plus experimental bridge plus animal mechanism — that this platform accepts as disease-relevant in other domains. Where it overreaches: when it drops the single-cohort, single-consortium fact; when one week of light data is treated as a life history; when the surrogate-endpoint experiment is described as if it showed disease; when the day-protective half is narrated as universal despite failing adjustment in the cardiovascular arm; and when the observational hazard ratios (1.3–1.5, honest but modest) are narrated like relative risks of 3 or 4.

Position B: the associations are substantially confounded and it is premature to name a new risk factor. Best evidence: reverse causation has a concrete, everyday pathway (the ill and the distressed are awake and lit, before diagnosis); UK Biobank's healthy-volunteer skew limits generalisability; the depression arm is cross-sectional; adjustment for socioeconomic position and sleep is always incomplete; the day-light protective associations in the cardiovascular analysis evaporated under fuller adjustment, which is exactly how a confounded night-light association would also behave; and every disease analysis so far shares one cohort and one analytic team, so the "multiple studies" appearance overstates the independence of the evidence. Where it overreaches: it has to explain away a dose-response gradient, polygenic independence, and a randomised experimental effect with the same wave of the hand; and its strongest form quietly reverts to the pre-2022 position — "light-at-night disease claims are wellness overclaims" — which was written about questionnaire-and-satellite studies and does not survive the sensor data unmodified.

The funding/bias dimension — cui bono, both ways. The unusual feature of this literature is that the core finding sells nothing: darkness is free, the research consortium is publicly funded, and no manufacturer profits when someone turns their lights off. That is a genuine credibility point and this platform says so. The residual interests are real but second-order. On the amplification side: the consortium has reputational stake in its cohort's findings; blue-blocker and "circadian lighting" vendors are already laundering these pattern findings into spectrum-product marketing they do not support; and media rewards "your lamp is giving you diabetes" framing. On the dismissal side: the lighting and screen industries benefit from the conversation staying on spectrum (where the consumer-harm evidence is genuinely null) and never arriving at pattern; and employers running night operations have no appetite for the gradient note. Neither set of stakes settles the science; the single-cohort caveat does more work than any of them.

Realised Position: We hold light at night as a Moderate-confidence, dose-responsive metabolic and disease risk marker with a plausible causal pathway partially demonstrated in humans — the strongest version of the claim the current evidence supports, and no stronger. We state the single-cohort ceiling in the same breath as the findings, every time. We act on it anyway, because the intervention is free, mechanistically coherent, harmless (falls aside), and independently supported by the sleep literature — the rare case where the expected value of acting under uncertainty is unambiguous. And we refuse the spectrum-product upsell in both directions: this evidence neither needs nor supports buying anything.

Cross-Pillar Connections

• darkness_and_sleep — owns darkness as a sleep-quality intervention and the sleep-architecture evidence; this entry supplies its metabolic hard-outcome anchors and the Mason causal receipt, and defers everything about sleep itself.
• shift_work_circadian_health — owns the occupational literature, its effect sizes, and mitigation for night workers; this entry cites the shift-work hazard once, for the gradient-coherence note, and defers all occupational detail.
• blood_sugar_regulation — owns glucose regulation as a system; this entry contributes light at night as an evidenced environmental contributor axis and defers all glycemic management.
• insulin_resistance_and_metabolic_dysfunction — owns insulin resistance as a condition, its measurement and reversal; this entry supplies the acute light-exposure experimental findings as one contributing exposure.
• circadian_rhythm_optimization — owns the full daily light-timing protocol, including the daytime bright-light half that ran protective in the mortality and mood analyses here; this entry owns only the night-exposure harm evidence.
• meal_timing_consistency_and_circadian_alignment — owns human meal-timing evidence; this entry's rodent mechanism (mistimed feeding under dim night light) is the animal end of that entry's thesis, and stays labelled rodent.
• sleep_foundations_for_baseline — owns the foundational sleep case; a dark bedroom serves both that entry's goals and this one's, which is why the intervention is doubly supported even under this entry's evidential uncertainty.
• blue_light_blocking_evening — owns the blue-blocker product evidence including the Cochrane null; this entry's findings are about light presence and timing and must not be converted into spectrum-product support.
• light_environment_and_health — the family hub; owns the tier map across all light claims, the cost-inversion doctrine, and the split firewall between spectrum claims (firewalled) and pattern claims (this entry).

What would change our mind

Industry bias note

Sources (10)

• Windred DP, Burns AC, Lane JM, et al. (2024). Personal light exposure patterns and incidence of type 2 diabetes: analysis of 13 million hours of light sensor data and 670,000 person-years of prospective observation. The Lancet Regional Health – Europe 42:100943. pubmed.ncbi.nlm.nih.gov/39070751↗/" target="_blank" rel="noopener">PMID 39070751↗. (Academic, publicly funded consortium (Monash/Flinders/Harvard lineage); no product interest — darkness is free; the consortium's stake is reputational.) 84,790 UK Biobank participants, one week of wrist-sensor light data, prospective follow-up: incident type 2 diabetes hazard ratios 1.29, 1.39 and 1.53 across brighter-night exposure groups, dose-dependent and independent of polygenic risk.
• Windred DP, Burns AC, Lane JM, et al. (2024). Brighter nights and darker days predict higher mortality risk: a prospective analysis of personal light exposure in >88,000 individuals. PNAS 121(43):e2405924121. (Academic; same consortium and cohort — adds outcome breadth, not cohort independence.) Brighter nights predicted 15–34% higher all-cause mortality across exposure groups and adjustment models, with the brightest tenth carrying 21–34%; brighter days were protective.↗
• Burns AC, Windred DP, Rutter MK, et al. (2023). Day and night light exposure are associated with psychiatric disorders: an objective light study in >85,000 people. Nature Mental Health 1:853–862. (Academic; same consortium; cross-sectional design — the weakest arm of the arc and labelled so.) Night light associated with ~30% higher odds of major depressive disorder; day light with ~20% lower odds; 86,772 participants.↗
• Windred DP, Burns AC, Rutter MK, et al. (2025). Light exposure at night and cardiovascular disease incidence. JAMA Network Open 8(10):e2539031. (Academic; same consortium and cohort; verified against the primary record for this entry in August 2026.) 88,905 participants, mean 9.5-year follow-up: brightest tenth of nights versus darkest half predicted higher incidence of coronary artery disease (+23–32%), myocardial infarction (+42–47%), stroke (+28%), heart failure (+45–56%) and atrial fibrillation (+28–32%), ranges spanning minimally to fully adjusted models; daytime-light protective associations appeared only in minimally adjusted models and did not survive lifestyle adjustment.↗
• Mason IC, Grimaldi D, Reid KJ, et al. (2022). Light exposure during sleep impairs cardiometabolic function. PNAS 119(12):e2113290119. pubmed.ncbi.nlm.nih.gov/35286195↗/" target="_blank" rel="noopener">PMID 35286195↗. (Academic sleep laboratory, publicly funded; the finding recommends turning lights off and sells nothing.) n=20 healthy adults, one night of sleep under 100 lux versus <3 lux: higher next-morning insulin resistance in the lit condition, with elevated heart rate and sympathetic-dominant heart-rate variability during sleep and no difference in melatonin — the effect ran through the autonomic pathway, through closed eyelids, in sleeping participants.
• Cheung IN, Zee PC, Shalman D, Malkani RG, Kang J, Reid KJ (2016). Morning and evening blue-enriched light exposure alters metabolic function in normal weight adults. PLoS One 11(5):e0155601. (Academic; note the design manipulated bright blue-enriched versus dim light, so intensity and spectrum are confounded — it must not be cited as spectrum-specific evidence.) n=19 healthy adults, inpatient: three hours of evening blue-enriched light produced a higher post-meal glucose peak versus dim light.↗
• Fonken LK, Workman JL, Walton JC, et al. (2010). Light at night increases body mass by shifting the time of food intake. PNAS 107(43):18664–18669. (Academic circadian biology; rodent.) Mice under 5-lux dim night light gained weight versus dark-night controls despite equivalent intake and activity; the driver was food intake shifted into the rest phase, and phase-restricted feeding prevented the gain. Rodent mechanism only — never to be presented as human evidence.↗
• Shift work and incident type 2 diabetes meta-analysis (2024). BMC Endocrine Disorders, doi:10.1186/s12902-024-01808-w.↗ (Academic meta-analysis; occupational-health context.) Pooled hazard ratio approximately 1.30 for type 2 diabetes in shift workers — cited here solely as the occupational anchor for the gradient note; all detail defers to the shift-work entry.
• Sooriyaarachchi P, Jayawardena R, Pavey T, King NA (2022). Shift work and the risk of metabolic syndrome: a systematic review and meta-analysis. Obesity Reviews. (Academic.) The metabolic-syndrome literature concurs with the diabetes direction for shift work.↗
• Singh S, Keller PR, Busija L, et al. (2023). Blue-light filtering spectacle lenses for visual performance, sleep, and macular health in adults. Cochrane Database of Systematic Reviews CD013244. pubmed.ncbi.nlm.nih.gov/37593770↗/" target="_blank" rel="noopener">PMID 37593770↗. (Independent systematic review, no industry funding — cited here as the guard against spectrum-product laundering of this entry's findings.) 17 RCTs: blue-blocking lenses probably confer no eyestrain benefit; sleep evidence indeterminate.

Related entries