What the research says, with sources.

In healthy middle-aged adults, 28 days of oral urolithin A (1,000 mg/day) vs placebo gave a small rise in naive-like CD8+ T cells and higher CD8+ fatty acid oxidation capacity. Other registered primary readouts were null: CD8+ T memory stem cells and SCENITH mitochondrial dependence. These are immune-cell surrogate markers; clinical benefit was not tested.

In healthy adults, 14 days of nicotinamide riboside (NR) or nicotinamide mononucleotide (NMN), but not nicotinamide (Nam), raised whole-blood NAD+ about 2-fold vs placebo (primary endpoint); a gut-microbial conversion mechanism was supported only by ex vivo experiments.

In healthy recreationally active men, 6 weeks of ubiquinol (300 mg/day) did not improve severe-intensity exercise time to exhaustion (primary outcome) vs placebo; muscle-biopsy subsample showed lower leak respiration, suggesting better mitochondrial coupling efficiency (surrogate).

In older adults with amnestic MCI, 12 weeks of nicotinamide riboside (1,000 mg/day) roughly doubled blood NAD+ but did not improve cognitive function (primary outcome), total cerebral blood flow or blood pressure vs placebo.

In a 4-week double-blind, placebo-controlled trial in 50 middle-aged adults with obesity, urolithin A 1000 mg/day did NOT improve the primary outcome (brachial artery flow-mediated dilation) or any secondary peripheral endothelial measure. A secondary brain measure (fNIRS task-evoked prefrontal oxygenation response during a working-memory task) increased versus placebo, with no change in cognitive task performance. Null primary; the brain signal is a secondary surrogate finding needing replication.

In a small prospective biopsy study of peripheral artery disease (PAD) patients, calf-muscle mitochondrial respiration normalised to mitochondrial content changed 6 weeks after successful revascularisation in patients with intermittent claudication (a decline the authors read as normalisation), but showed no significant change in patients with chronic limb-threatening ischemia (CLTI), whose mitochondrial content marker (citrate synthase activity) rose. Observational, no control group, very small paired CLTI subgroups.

A systematic review of randomised controlled trials (RCTs) found only 3 small, short trials of coenzyme Q10 (CoQ10) in heart failure with preserved ejection fraction (HFpEF) and concluded the evidence is insufficient to tell whether CoQ10 helps: deaths and heart-failure hospitalisations were too rare to analyse, and the one trial measuring quality of life gave conflicting results between its two comparisons. All primary-outcome evidence was rated very low certainty (GRADE). This is a null/uninformative review, not evidence of benefit or harm.

Landmark T01 #1, main finding (CELL AND MOUSE WORK ONLY; NOT HUMAN PROOF). In cultured human cancer cell lines, death caused by excess copper (delivered by copper ionophores such as elesclomol, or by over-expressing the copper importer SLC31A1) did not behave like apoptosis, ferroptosis or necroptosis: blockers of those pathways did not prevent it and it continued in BAX/BAK-knockout cells (Fig 1; Fig 6G). It depended on mitochondrial respiration - cells relying on respiration were described as 'nearly 1,000-fold more sensitive' than glycolytic cells, and electron-transport-chain inhibitors and 1% oxygen reduced the killing (Fig 2). Genome-wide CRISPR screens found that losing FDX1 or protein-lipoylation genes (LIAS, LIPT1, DLD, DLAT, PDHA1, PDHB) protected cells (Fig 3). Proposed mechanism: copper binds lipoylated TCA-cycle proteins (DLAT, DLST), DLAT clumps (oligomerises), iron-sulfur cluster proteins are lost and proteotoxic stress follows (Figs 5-6). The authors named this 'cuproptosis'. Liver from Atp7b-knockout mice (a Wilson-disease copper-overload model) showed the same protein changes (Fig 6K). No human was treated in this study.

Landmark T01 #1, methods and limits (CELL AND MOUSE STUDY - NOT A HUMAN TREATMENT). The cuproptosis model rests on human cancer cell lines, two single-line CRISPR screens, cell-based biochemistry and one mouse-liver experiment, with no human intervention. Main limits: (1) copper was mostly forced in with ionophores (elesclomol, disulfiram/cupric-DDC, NSC319726) and the FDX1/lipoylation dependence disappears above ~40 nM elesclomol (fig S4L-P); (2) each CRISPR screen used a single line (OVISE whole-genome; A549 metabolism-gene); (3) mouse evidence is protein changes in Atp7b-/- liver, not a survival or treatment outcome; (4) human tumour data are FDX1-vs-lipoic-acid staining correlations; (5) the paper carries an erratum. One true internal inconsistency (tumour count 208 vs n=57) plus a free-copy-vs-publisher COI contradiction are recorded below.

Landmark T01 #1, what came after (1): mechanism follow-ups (still not human proof). Three primary studies I read in full confirmed and refined the FDX1->lipoylation step, all in human cell lines and yeast only (none did new zebrafish or mouse work). (1) Dreishpoon MB et al., J Biol Chem 2023 (PMC10462841): FDX1 directly binds the lipoyl-synthase enzyme LIAS and promotes lipoylation rather than acting only through iron-sulfur-cluster biosynthesis; LIAS protein is actually stabilised (rises), not lost, when FDX1 is lost; FDX1 loss causes loss of respiration and sensitivity to mild glucose starvation. (2) Joshi PR et al., J Biol Chem 2023 (PMC10470009): FDX1 (not FDX2) is required for lipoic-acid synthesis, but FDX1-knockout cells grow under 1% oxygen without lipoylation being rescued, arguing against a hypoxic salvage pathway. (3) Zulkifli M et al., PNAS 2023 (PMC10013847): FDX1 reduces elesclomol-bound copper inside mitochondria to release it to cytochrome c oxidase, with a separate FDX1-independent release outside mitochondria. These extend the landmark's mechanism; none is a human treatment study.

Landmark T01 #1, what came after (2): translation and the literature explosion. Since 2022 'cuproptosis' is a large research field, but direct human evidence is thin. OpenAlex (retrieved 2026-10-18) gives about 4,700-4,900 citing works (cites:W4221115671 filter = 4,749; the record's cited_by_count = 4,872); title.search 'cuproptosis' = 2,332, 'cuproptosis signature' = 370, 'cuproptosis prognostic' = 301 - overwhelmingly bioinformatic gene-signature papers, not experiments. A ClinicalTrials.gov search for 'cuproptosis' (2026-10-18) returns only two studies: NCT07843537 (interventional; MINTER phase 2 open-label, disulfiram 250 mg + copper gluconate 2 mg + radiotherapy in thoracic cancer, not yet recruiting, est. start 2027-03-01, first submitted 2026-09-21, explicitly invoking cuproptosis) and NCT06833918 (observational, acute myocardial infarction, not yet recruiting). No trial has yet tested cuproptosis as a treatment mechanism; note copper-ionophore drugs were tested in people BEFORE the term existed (e.g., the elesclomol + paclitaxel phase 3 melanoma trial SYMMETRY, negative overall, cited in the landmark's Discussion). Concrete follow-ups I read in full: Sun L et al., Nat Commun 2023 (PMC10589265), links copper/cuproptosis to gastric cancer using 48 paired human tumour/normal tissues + cell and mouse-xenograft experiments; and Tao Z et al., Medicine (Baltimore) 2024 (PMC11596713), a small retrospective human study in 76 Wilson-disease patients. So four years on, cuproptosis in humans rests on observational and bioinformatic work.

Randomized crossover, single acute session, unblinded (lights-off) sham: one 12-min whole-body red/near-infrared LED session raised resting energy expenditure from pre to post in women with obesity, with no change in substrate use. Normal-weight women also rose, so the effect is not shown to be obesity-specific. Acute pilot response only; mitochondria not measured.

In adults with chronic upper-trapezius myofascial trigger points, 808 nm laser at 16 J per point (8 sessions) was not superior to sham for pain (primary outcome); neck disability (secondary) improved more with laser.

In a single-blind, sham-controlled multicentre trial partly authored by the device maker, a home-use 630+850 nm LED patch plus rehabilitation produced greater 28-day pain improvement in chronic non-specific low back pain than a low-power red sham plus rehabilitation. The function (JOA) difference was below the clinically important threshold.

In an open-label, assessor-blinded RCT of 64 young patients after gingivectomy, a 940 nm diode-laser photobiomodulation protocol produced slower early wound epithelialisation and slightly higher early pain than a hyaluronic acid (HA) spray; 6-week gum outcomes were similar. There was no untreated or sham arm, so this does not show whether either treatment beats natural healing.

A meta-analysis of 28 RCTs (3573 children, all but one in China) found that repeated low-level red-light (RLRL) therapy, a form of photobiomodulation using 650 nm light, slowed myopia progression over 12 months compared with control: spherical equivalent refraction worsened about 0.68 dioptres less and the eye grew about 0.30 mm less in axial length. Certainty was moderate for refraction and low for axial length (GRADE). Safety assessment was limited: no severe adverse events were reported, but retinal structure was too inconsistently reported to pool, and evidence beyond 1 year was not analysed.

In a manufacturer-sponsored, double-blind, sham-controlled RCT of 64 adults with painful diabetic peripheral neuropathy (DPN), 3 weeks of twice-daily home use of a dual-wavelength (640 nm red + 405 nm violet) low-level laser device on the feet gave more people at least a 30% drop in foot pain than sham (72.7% vs 32.3%, intention-to-treat with imputation). Short-term pain outcome only (to week 7), small trial, and the active arm guessed its allocation above chance.

A systematic review of abdominal photobiomodulation (PBM, light applied to the belly) and the gut-brain axis found only 9 studies: 5 small human studies (1 pilot RCT, case series, a single case report, and overlapping reports from one 12-patient Parkinson's cohort) and 4 animal studies. The only RCT (53 people with Alzheimer's disease, head plus abdomen device) did not improve its primary cognitive score. The evidence is early and mostly preclinical; the review's own 'favorable recommendation' wording goes beyond its data.

An umbrella review of 100 systematic reviews of low-level light therapy / photobiomodulation found that 62 reviews concluded in favour of the intervention, 34 were neutral/mixed, and 4 unclear, with the highest share of favourable conclusions in oncology-related complications (especially oral mucositis). 90 of 100 included reviews scored low or critically low on AMSTAR 2, some reviews overlapped heavily in the primary studies they counted, and the umbrella did not compute pooled effect sizes or GRADE certainty, so this maps the review literature rather than proving efficacy.

In a single-centre, randomised trial in Brazil of children and teens (1-18 years) receiving high-dose methotrexate (HDMTX) chemotherapy, preventive photobiomodulation (PBM, low-level red laser) applied from outside the face (extraoral, about 30 s per session) and inside the mouth (intraoral, about 6 min per session) were followed by oral mucositis in 30.5% vs 41.1% of children, a difference that was not statistically significant. The trial had no untreated control group and was designed and powered to detect a difference, not to prove equivalence, so it does not show that the two methods are equally effective or that PBM itself prevented mucositis; it was also not prospectively registered.

A systematic review and meta-analysis of 23 randomised trials of low-level (LLLT) and high-intensity (HLLT/HILT) laser therapy for rotator cuff-related shoulder pain reported a reduction in overall pain, but the forest plot shows this effect is a standardised effect size (Hedges's g -5.22, I2 98.5%) driven by several trials with implausibly small standard deviations, not a 5-point drop on a 0-10 pain scale as the text's 'MD' label implies. Pain at rest, during movement and at night, and all four function scores were not significantly improved, and no certainty (GRADE) rating was done, so this review does not establish that laser therapy helps rotator cuff pain.

Landmark T02 #1, main finding. In a 12-month multicentre randomised controlled trial in Chinese children aged 8-13 with low-to-moderate myopia, adding repeated low-level red-light (RLRL) therapy - a desktop device emitting 650 nm red light, 3 minutes per session, twice a day, 5 days a week - to single-vision spectacles (SVS) slowed eye growth over one year. Adjusted 12-month axial elongation (the pre-registered primary outcome) was 0.13 mm with RLRL vs 0.38 mm with SVS; between-group difference 0.26 mm (95% CI 0.20-0.31; P<0.001), a 69.4% reduction. Myopia (SER) progressed -0.20 D vs -0.79 D, difference 0.59 D (95% CI 0.46-0.72; P<0.001), a 76.6% reduction. No severe adverse events, no loss of best-corrected vision, and no OCT structural damage were observed among the children scanned.

Landmark T02 #1, methods and limits (registry and funding/COI check). This RCT (NCT04073238) pre-registered axial length as its single primary outcome, with SER and other biometry/visual acuity as secondaries; the current registry record and protocol v.7 (May 2019) match the paper on these. Registration was retrospective by about 26 days (first submitted 2019-08-18; actual start 2019-07-23). The device maker (Suzhou Xuanjia Optoelectronics / Eyerising) provided devices free and, per the paper, gave no research funding; its system also ran compliance monitoring. The paper discloses a device patent for Z.Z. and M.H.; M.H.'s later papers disclose a director/equity role in Eyerising. The biggest design weakness is the absence of a sham control.

Landmark T02 #1, what came after (1): replication and a sham-controlled trial. Two primary follow-ups I read in full. (1) Dong J, Zhu Z, Xu H, He M, Ophthalmology 2023 (PMID 36049646) - the first double-masked, SHAM-controlled RLRL RCT; 112 Chinese children 7-12, 6 months, same Eyerising device at 100% vs a sham at 10% power: the 100%-power group beat sham on the primary outcome (SER change +0.06 D vs -0.11 D, P=0.003) and on axial length (+0.02 mm vs +0.13 mm, P<0.001), so benefit survives masking - but the 10%-power 'sham' itself progressed less than historical spectacle controls, hinting even low-power light may act. (2) Xiong R et al., Clin Exp Ophthalmol 2022 (PMC10086781) - the 2-year post-trial, NON-randomised follow-up of the NCT04073238 cohort: continued RLRL kept the best control over 2 years (axial 0.16 mm vs 0.64 mm for spectacles-only, 75.0% efficacy), but STOPPING RLRL produced a rebound - second-year axial elongation was 0.42 mm, FASTER than the 0.28 mm seen the same year in never-treated children (authors call it a 'modest rebound'; 2-year cumulative RLRL-SVS 0.50 vs SVS-SVS 0.64 mm).

Landmark T02 #1, what came after (2): the safety question. Independent work has raised retinal-safety concerns the original trial's OCT/BCVA checks could miss. (1) Ostrin LA & Schill AW, Ophthalmic Physiol Opt 2024 (PMC10922340; read in full) measured two commercial red-light myopia devices (the Sky-n1201a and Future Vision; it did NOT test the Eyerising) and calculated that 3 minutes of continuous viewing can approach or exceed the retinal maximum permissible exposure (MPE); recomputing the landmark's Eyerising device (0.29 mW/4-mm pupil -> 2.31 mW/cm2 -> 0.889 mW over a 7-mm aperture > Class-1 MPE 0.385 mW). It also recounts (a single-patient case report, as described by Ostrin & Schill 2024; original not read - not open access) a 12-year-old whose BCVA fell from 6/6 to 6/9 after 5 months of use with bilateral foveal ellipsoid-zone disruption and 'improved to 6/7.5 three months after' therapy stopped. (2) Liao X et al., JAMA Ophthalmol 2025 (PMC12022868; read in full): among RLRL users vs non-users, paracentral cone density was lower (at 0.3 mm temporal, -2.1 x10^3 cells/mm2; 95% CI -3.68 to -0.59; one-sided P=.003), and 'abnormal low-frequency, high-brightness signals' (drusen-like) near the fovea were more common in users (reported OR 7.23, not reproducible from the paper's printed eye counts: 7/97 vs 1/74 gives OR about 5.7, two-sided Fisher P about 0.14). A published dispute (letter + reply, 2026; abstract/record only) and a 2026 4-device re-evaluation (abstract/record only) have since followed; a 3-year real-world cohort (abstract/record only) reports no BCVA loss and no treatment-duration-dependent full-field ERG change, with a minimal, reversible OCT change in 4 eyes (abstract only); its senior and corresponding author, Mingguang He, is a director and shareholder of Eyerising and a named inventor on related patents (BJO disclosure). Long-term retinal safety is unresolved.

In women receiving chemotherapy for breast cancer, 6 weeks of morning bright light therapy via light glasses did not improve insomnia or fatigue (dual primary outcomes) vs non-light groups. Brief CBT-I improved insomnia by the end of the 6-week intervention, with partial rebound during follow-up.

In UK Biobank adults, higher wrist-sensor-measured daytime light exposure was associated with modestly lower incident dementia risk over about 8 years (HR 0.84; crude absolute difference about 0.05 percentage points). Nighttime light was not associated. Association only, not causation.

A systematic review and meta-analysis of bright light therapy (BLT) to prevent delirium in hospitalised adults (8 studies, 1190 participants, mostly intensive care) found an uncertain effect: pooled risk ratio 0.64 (95% CI 0.38 to 1.07), which includes no effect, with very low certainty (GRADE). The largest trial (734 ICU patients, ceiling lighting) found no reduction; a benefit appeared only when that trial was excluded, which is a sensitivity analysis, not the main result.

In an exploratory longitudinal EEG study of 16 male Antarctic winter-over expeditioners, delta-band eyes-closed minus eyes-open (EC-EO) log-power differentiation shifted progressively more negative across the mission, with the main change between pre-polar-night and early post-polar-night assessments and persistence into austral summer. This is a candidate resting-state EEG reactivity index (the authors deliberately avoid the word biomarker), not a clinical outcome. There was no control group and no light, sleep, mood or melatonin data, so the change cannot be attributed to polar-night light deprivation rather than isolation, season, workload or time on the mission.

Landmark T03 #1, main finding (revised against the author accepted manuscript; publisher version paywalled - accepted manuscript + free supplement + nature.com public back matter read). In 86,772 UK Biobank adults (aged 62.4 +/- 7.4 y, 57% women) who wore a wrist light sensor for one week, more night-time light was associated with HIGHER odds of most psychiatric outcomes and more daytime light with LOWER odds of several, after adjustment. The paper's PRIMARY model compares the brightest quartile (Q4) with the darkest (Q1) referent (Fig 2, Tables 2-3; fully adjusted Model 3: age, sex, ethnicity, photoperiod, employment, physical activity). Night-light Q4-vs-Q1 adjusted odds ratios (aOR): major depressive disorder 1.30 (95% CI 1.23-1.38), PTSD 1.34 (1.22-1.48), generalized anxiety disorder 1.23 (1.11-1.36), self-harm 1.27 (1.14-1.42), psychosis 1.21 (1.09-1.34); for BIPOLAR disorder the overall (omnibus) night-light test was NON-significant (omnibus chi2 P=0.18) and only the top quartile was elevated (1.20, 1.02-1.42) - so the abstract's listing of bipolar among night-light associations rests on the top quartile only. Day-light Q4 aOR: major depression 0.81 (0.76-0.87), psychosis 0.69 (0.61-0.79), self-harm 0.76 (0.67-0.87), PTSD 0.82 (0.73-0.92); day light was NOT associated with generalized anxiety (omnibus chi2 P=0.61) or bipolar disorder (omnibus chi2 P=0.22). This is a HUMAN CROSS-SECTIONAL (observational) study: lifetime psychiatric case status was compared with light measured on average 1.86 years EARLIER, so it shows ASSOCIATION, NOT causation - the authors state 'we acknowledge the possibility of reverse causation and that longitudinal studies will be needed to establish the temporality of the associations we observed.' The authors suggest 'avoiding light at night and seeking light during the day may be a simple and effective, non-pharmacological means of broadly improving mental health' - their interpretation, not demonstrated by this design. No medical advice.

Landmark T03 #1, methods and limits (revised against the author accepted manuscript; publisher version paywalled - accepted manuscript + free supplement + nature.com public back matter read). Design: cross-sectional secondary analysis of UK Biobank (no trial registration; observational). Light measured with a wrist-worn Axivity AX3 accelerometer / APDS9007 silicon photodiode (spectral sensitivity ~470-650 nm), dominant wrist, 7 days (2013-2015); psychiatric outcomes from the online Mental Health Questionnaire completed on average 1.86 years LATER (2016-2017). Device output was converted to approximate lux and piece-wise calibrated against a Tektronix J17 photometer in a validation run on only 7 AX3 devices from the original study, tested at 34 light levels from 0 to 5,050 lux, ~8 years after the recordings (recovered r2=0.82 low / 0.81 high, attenuated/non-linear at the low end). Within-person light ICC was 0.82 (95% CI 0.81-0.83), based on n=2,988 participants who repeated actigraphy four times, on average 3.23 y after the main assessment. Daily 24-h profiles were reduced by factor analysis to 'day light' (07:30-20:30) and 'night light' (00:30-06:00), each split into four EQUAL-sized quartiles (day-night correlation rs=0.10). PRIMARY analysis: logistic (disorders) / linear (symptom scales) regression of each quartile vs the Q1 referent, with likelihood-ratio chi2 omnibus tests and three nested models (Model 3 = age, sex, ethnicity, photoperiod, employment, physical activity). An additional interaction model tested the day x night interaction using per-SD log10 light (Supplementary Tables 7-8); the paper does not call this 'secondary' - the quartile models are what it calls its primary findings/hypotheses. Results were robust across sensitivity analyses (excluding shift workers; excluding clinical subgroups; adjusting for sleep [Model 4], urbanicity [Model 5] and cardiometabolic health [Model 6]), with author-noted exceptions: after sleep adjustment (Model 4) the night-psychosis omnibus test was no longer significant (P=0.09; Q4 1.12, 1.01-1.25; Suppl Table 12), and after cardiometabolic adjustment (Model 6) the day-PTSD omnibus test was no longer significant (P=0.11; Q4 0.87, 0.78-0.98; Suppl Table 16) - the authors describe day-PTSD as 'weaker'. Bipolar Q4 was also not significant in Models 4 and 6 (Suppl Tables 12, 16). The clinical-subgroup exclusion applied to the symptom scales only (Suppl Table 11). Author-stated limitations (Discussion): cross-sectional (possible reverse causation; longitudinal needed); wrist (not ocular) light is 'a coarse estimate'; actigraphy and outcomes not simultaneous; device coverage cannot be distinguished from true darkness; the photopic APDS9007 'cannot report melanopic equivalent daylight illuminance'; lux from a late device validation. No medical advice.

Landmark T03 #1, what came after (1): the same team extended the day/night light framework from a cross-sectional mental-health association to a PROSPECTIVE hard outcome - death. Windred et al. 2024 (PNAS; read in full) followed 88,905 UK Biobank adults who wore the same wrist light sensor for one week, for a mean 8.0 years (3,750 all-cause and 798 cardiometabolic deaths). Using Cox proportional-hazards models for all-cause mortality and competing-risks proportional subhazards models for the cause-specific (cardiometabolic) outcomes, adjusting day and night light together (percentile groups 0-50 [ref], 50-70, 70-90, 90-100) plus age, sex, ethnicity, photoperiod and socioeconomic/lifestyle factors (Model 3): the brightest-night group (90-100th percentile) had HIGHER all-cause mortality (adjusted HR 1.21, 95% CI 1.08-1.35, P=0.00064) and higher cardiometabolic mortality (subhazard HR 1.33, 1.06-1.68, P=0.0069), while the brightest-day group had LOWER all-cause mortality (aHR 0.83, 0.72-0.94, P=0.0042); effects were dose-dependent. A computational model of the brain's circadian pacemaker showed that higher modelled circadian amplitude was linked to lower all-cause mortality (mean amplitude aHR 0.96 per SD higher, 0.92-0.99), and early or late circadian phase predicted higher mortality. Adjusting for SHORT sleep attenuated the night-light association (90-100th aHR 1.11, 0.99-1.24, no longer significant), which the authors say indicates short sleep 'partially explains' the night-light association (the 70-90th night group stayed significant: 1.11, 1.02-1.22). This is still OBSERVATIONAL/correlational - the authors state 'since this is a correlational study, it is possible that light exposure patterns and premature mortality are explained by other unmeasured factors'; association, not proof of cause. No medical advice.

Landmark T03 #1, what came after (2): a group with NO author overlap with the landmark (though Wallace, Redline and Sofer sit in the same Brigham and Women's Division of Sleep and Circadian Disorders / Harvard Division of Sleep Medicine as landmark co-authors Burns, Saxena and Lane - authorship independent, institution not), in a different population, device and metric, tested the day-light-and-mood half of the finding; the letter cites the landmark (ref 3). Wallace et al. 2024 (JAMA Network Open research letter; read in full) analysed 6,639 US adults in NHANES 2011-2014 who wore a wrist ActiGraph GT3X+; bright-light exposure was 'time above 1000 lux' (TALT1000, hours/day) and depression was the PHQ-9. In the fully adjusted model (age, sex, race/ethnicity, season, employment, physical activity, sleep duration, BMI, serum cotinine): more time in bright light (>1000 lux) was associated with LOWER depression symptoms (linear beta = -0.19 PHQ-9 points per hour, 95% CI -0.34 to -0.04, P=0.02), consistent in direction with the landmark's day-light/depression association, in a different cohort (different design, exposure and outcome, so not a like-for-like replication). Two nuances temper it: (1) the association was NOT significant for more severe depression (PHQ-9 >=10: OR 0.95, 0.83-1.08, P=0.39); (2) after additionally adjusting for sleep regularity it fell to non-significant (beta = -0.11, -0.26 to 0.04, P=0.13), while more bright light predicted more-regular sleep (beta 1.60, 0.95-2.25) and more-regular sleep predicted lower depression (beta -0.05, -0.06 to -0.03) - so the authors conclude 'greater sleep regularity partly explained the association between greater bright light exposure and lower depression symptoms.' Exploratory vitamin D was linked to light and to sleep regularity but NOT to depression. CROSS-SECTIONAL: the authors say 'causality cannot be ascertained, and associations may be bidirectional' (mood may drive time outdoors). Association, not causation; no medical advice.

For mild-to-moderate actinic keratoses (an approved PDT indication), one session of simulated-daylight ALA-PDT cleared fewer lesions at 1 year than conventional red-light PDT and failed non-inferiority, but was far less painful.

A systematic review and meta-analysis of 15 observational studies found that phototherapy for newborn jaundice was associated with 24% higher odds of a later cancer diagnosis (up to age 31); the association was strongest for blood cancers (OR 1.40) and solid-organ cancers were borderline (OR 1.18, lower CI 1.01). Because childhood cancer is rare, the authors estimate this would mean roughly 51 extra cancers per 100,000 treated children (an approximate illustration), and they stress it must be weighed against the serious harms of untreated high bilirubin. Association only; several studies did not adjust for confounders such as prematurity or bilirubin level.

In a phase 2 proof-of-concept RCT in adults with extensive active non-segmental vitiligo, baricitinib 4 mg/day plus narrowband UV-B did NOT meet its prespecified primary endpoint: the mean week-36 reduction in total VASI (-44.8%; 95% CI -58.4% to -31.3%) was not significantly greater than the prespecified 42.9% repigmentation threshold taken from historical narrowband UV-B-alone data. The positive signal vs the concurrent placebo+NB-UVB arm (-44.8% vs -9.2%; P=.02) came only from a post hoc, 2-sided analysis.

Methods and limits of the baricitinib+NB-UVB vitiligo trial: a small (n=49) noncomparative phase 2 RCT funded by Eli Lilly, prospectively registered (NCT04822584) with a single primary endpoint (total VASI at week 36) whose success criterion was beating a 42.9% historical NB-UVB threshold rather than the concurrent placebo arm. Several registered secondary outcomes (all week-48 measures and the blood/skin inflammatory biomarkers) are not reported in this paper, and the article carries several internal counting inconsistencies.

A systematic review and meta-analysis (volume dated 2026; online 27 Dec 2025) of JAK inhibitor plus narrowband UV-B in adult non-segmental vitiligo (4 studies, 217 adults, of which only 1 is an RCT - this landmark - the other 3 non-randomized) found combination therapy pooled a greater total-VASI reduction (MD -4.96; 95% CI -9.29 to -0.63; p=0.02) but with very high heterogeneity (I2=85%). Its direction is qualifiedly consistent with the landmark; it is NOT an independent replication because it incorporates the landmark's own data.

A 2025 narrative (non-systematic) review of the vitiligo treatment pipeline, from an author group independent of the landmark, situates the baricitinib+NB-UVB trial among JAK-inhibitor studies. Its prose accurately states the week-36 VASI result (-44.8% vs -9.2% for placebo), that this did not surpass the prespecified 42.9% threshold, and that the between-group difference came from a post hoc analysis; however the review's clinical-trials table mislabels the same trial. The review frames combination therapy as, in its words, 'likely to remain the most promising approach' and lists tissue-resident memory (TRM) and regulatory T-cell targeting among emerging directions.