Red Light Therapy Devices: What the Evidence Supports and How to Read a Dose

Summary

Consumer red/near-infrared light devices ("red light therapy", photobiomodulation) are a real product category with wildly uneven evidence that must be read per indication, not as a category: hair regrowth in androgenetic alopecia is the genuine best claim (a meta-analysis of eleven double-blind sham-controlled trials, though manufacturer-sponsored and FDA-cleared, not approved); skin photoaging holds Moderate support from panel-style trial protocols that masks then borrow across a form-factor gap; muscle recovery is Moderate at explicitly low certainty; the morning-eye and glucose findings ar

Why Moderate

The entry holds at Moderate as its centre of gravity, because that is where its two best consumer indications genuinely sit and because the category must not inherit its single best sub-claim.
• Why not the tier above (Strong): the category's one Strong-calibre claim (hair) carries a structural sponsorship caveat and does not generalise — skin declines to pool, muscle is low-certainty, and everything systemic is Emerging or below. A category tier of Strong would launder eleven hair trials across a thousand-dollar product aisle.
• Why not the tier below (Emerging): eleven double-blind sham-controlled RCTs in one indication plus real controlled skin trials is materially better than Emerging; rounding the category down would misstate the hair evidence and forfeit credibility on the honest parts.
• Per-sub-area split: hair/AGA = Strong Evidence (manufacturer-sponsored, FDA-cleared-not-approved) · skin photoaging = Moderate (panel protocols; masks unproven as sold) · muscle recovery = Moderate at low certainty · eyes (morning 670nm) = Emerging (single lab, surrogate outcome) · glucose = Experimental (n=30, unreplicated) · NIR-deficiency / "light nutrition" = Experimental hypothesis, firewalled · daylight dose arithmetic = Foundational physics (with the sunrise figures explicitly calculated, not measured).

Practical takeaway

Start from the free source. Mid-morning to mid-afternoon daylight delivers roughly 20–35 mW/cm² across the red/NIR band — inside the therapy's own preferred intensity window, over your whole body, at zero cost. If the general "red light is good for you" claim is what attracts you, daylight already covers it, with the one honest coupling stated plainly: midday NIR arrives with midday UV, so ordinary sun-sense (duration, skin type, no burning) applies. Sunrise and sunset walks remain excellent — for circadian anchoring and near-zero UV — but do not count them as a red-light dose; the in-band photons at low sun are one to two orders of magnitude below the therapeutic window.

If you are considering a device, buy per indication, not per category:
• Hair (androgenetic alopecia): the one indication with genuinely strong trial support and no free substitute. FDA-cleared helmets/combs are a defensible purchase for someone who wants a non-drug option or an adjunct — with realistic expectations (absolute gains around +14–19 hairs/cm² in trials, not restoration) and the sponsorship caveat priced in. Compare against the established drug options in the hair-loss literature before spending.
• Skin photoaging: the trial protocols ran 10–30 sessions over four to fifteen weeks on panel-class devices. If you buy, buy the form factor and commit to the protocol the evidence used — a mask used sporadically at 1–3 J/cm² per session is not the thing the trials tested.
• Muscle recovery: plausible, modest, low-certainty. Reasonable for a serious trainee who already has sleep, protein and programming handled; irrational as a first purchase (see the recovery-fundamentals entries).
• Eyes and glucose: do not buy for these. Single-lab Emerging and unreplicated Experimental respectively; the eye finding is morning-only and surrogate; the glucose finding's honest absolute effect is 7.3%.

Reading a spec sheet: demand irradiance at usage distance (not 0mm), treat advertised numbers as ±3× until independently measured, and compute your session fluence (irradiance in mW/cm² × seconds ÷ 1000 = J/cm²) against what the trials for your indication actually delivered. If the seller cannot tell you the trial protocol they are matching, that is the answer.

Eye protection is non-negotiable on any high-irradiance device near the face, and treat any mask with blue-wavelength emitters as a different, higher-risk product class (see Risks).

Evidence detail

Why This Entry Exists

Red light therapy is one of the fastest-growing consumer wellness categories — panels, masks, helmets, belts, wands — with price tags running from $30 to well past $1,000, and a marketing engine that quotes real studies far beyond what they showed. The category is unusual in that the underlying science is not fake: photobiomodulation has decades of literature, a plausible mechanism, and at least one indication with a genuinely respectable randomised-trial base. That partial legitimacy is exactly what makes it hard for a user to navigate — the honest answer is neither "it's all placebo" nor "buy the panel", and almost nobody selling content on the topic has an incentive to say so.

The structural fact this entry exists to state is the cost inversion that runs through the whole light-and-health family: every light intervention with Strong evidence behind it — outdoor time, morning light, a dark bedroom, daylight itself — costs nothing, while nearly everything priced over $100 sits at Emerging or Experimental. The one partial exception is the hair-loss device class this entry documents, and even its Strong-calibre evidence arrives with a manufacturer-sponsorship caveat attached. Money in this category flows toward the weakest-evidenced tier. A user who understands that single fact, plus the dose arithmetic that lets them evaluate any device claim, is protected against most of what the category sells.

The second job is the buyer's-guide machinery itself: what irradiance and fluence numbers mean, why advertised specs cannot be trusted at face value, why a mask and a panel are physically different propositions, and why the sun — at the right hours — is the incumbent red/NIR source that no trial has ever beaten head-to-head, because no such trial exists.

What bad advice this protects against, in all directions:
• "Red light therapy is proven — the science is settled, get a panel" → the evidence is per-indication and mostly Moderate or below; the best claim (hair) is manufacturer-sponsored; skin protocols ran committed multi-week panel regimens of 10–30 sessions that a casual mask habit does not reproduce; "proven" flattens exactly the tier structure a buyer needs.
• "It's all pseudoscience and placebo" → also wrong. Eleven double-blind sham-controlled RCTs on hair density is not placebo-tier evidence, and dismissing the whole category makes the platform untrustworthy on the parts that are real.
• "Watch the sunrise for your red light dose — golden hour is nature's red light therapy" → golden hour is red-rich in proportion but photon-poor in dose: roughly 0.5–3 mW/cm² in-band versus ~20–35 at midday (figures calculated from standard atmospheric physics, not measured spectral datasets). Sunrise's real credit is circadian phase and near-zero UV. The fusion of the two claims sells a dose that is not there.
• "You have an infrared deficiency from living indoors — modern lighting is starving your mitochondria" → a self-labelled hypothesis whose lead author sells NIR lighting; three unproven links stacked; stays firewalled as such.
• "A study showed red light cuts blood sugar 27.7%" → the famous figure is the incremental glucose area-under-curve in one unreplicated n=30 study; the absolute reduction was 7.3%, the peak reduction 12.1% — and the study used an LED, so it also cannot be recruited for "LED red light is inferior or harmful".
• "Red light lowers blood pressure" → citation laundering: the systemic blood-pressure finding in circulation is a UVA study, routinely re-quoted as if it were a red-light result.

Evidence

Read the tiers, not the thesis. This category's literature spans a genuinely well-replicated trial base at one end and single-lab surrogate findings at the other, with the marketing deliberately blending them. Every claim below is labelled by evidence type.

The indications, tier by tier:

1. Hair — androgenetic alopecia: the category's best claim (Strong Evidence, with a sponsorship caveat). HUMAN RCT meta-analysis: a 2019 meta-analysis (Lasers in Medical Science, PMID 30706177) pooled eleven double-blind, sham-device-controlled randomised trials of low-level laser/light therapy for androgenetic alopecia and found a large standardised effect on hair density (SMD 1.316 versus sham), with absolute gains on the order of +14–19 hairs/cm². Two honest caveats travel with it: most constituent trials were manufacturer-sponsored, and the devices are FDA-cleared (a 510(k) equivalence pathway) — not FDA-approved, a distinction marketing reliably erases. Cui bono: the trial base was substantially funded by the companies selling the helmets and combs; the effect direction has nonetheless survived double-blind sham control eleven times, which is more than most of the wellness aisle can say. This is also the one indication on the list with no free substitute — daylight on the scalp is not a controlled AGA protocol.

2. Skin — photoaging: real trials, borrowed by the wrong form factor (Moderate). HUMAN RCT: controlled trials exist — Wunsch & Matuschka 2014 (Photomedicine and Laser Surgery, n=136, randomised, four arms including a placebo-equivalent control; improved skin complexion and intradermal collagen density measures) and a randomised periocular wrinkle trial (Mota et al. 2023, PMID 36780572 — split-face, red versus amber light, both arms active, no sham arm) — using multi-week panel-class protocols of roughly 10–30 sessions (Wunsch: twice weekly for fifteen weeks; the periocular trial: ten sessions in four weeks). Systematic reviews (e.g. PMC11049838) acknowledge the signal but decline to pool, citing heterogeneity of devices, doses and endpoints. Cui bono: mask manufacturers quote this panel-protocol literature to sell a different device class delivering a different dose — the evidence is borrowed across a form-factor gap the ad never mentions.

3. Muscle recovery and pre-conditioning (Moderate, at explicitly low certainty). HUMAN RCT meta-analyses: pooled analyses of photobiomodulation for delayed-onset muscle soreness and pre-exercise conditioning (including a 2024 meta pooling 34 RCTs, and PMC12286287) report moderate effects on soreness and some performance measures — with certainty graded low, small trials, and a notably narrow, partly device-linked authorship cluster producing much of the literature. Cui bono: several prolific authors in this sub-field have device-company affiliations; independent groups find smaller or null effects.

4. Eyes — morning 670nm and ageing colour vision (Emerging; single lab, surrogate outcome). HUMAN experimental, small: Shinhmar et al. (Scientific Reports 2021;11:22872) exposed 20 adults aged 37–70 to a single three-minute 670nm exposure between 8 and 9 a.m. and reported a ~17% average improvement in cone-mediated colour contrast sensitivity at three hours, persisting a week in the ten participants retested; a six-person afternoon-exposure arm months later showed no effect — the time-of-day contrast serving as the study's internal control condition rather than a sham-device arm. An earlier study from the same group (Journals of Gerontology A 2020, PMID 32596723) reported related findings. This is one laboratory, small samples, a surrogate visual-function outcome, and no independent replication. Cui bono: no direct commercial stake identified for the lab, but the finding is heavily merchandised by third-party device sellers who omit the morning-only condition and the sample size.

5. Glucose — the honest framing of a famous number (Experimental). HUMAN experimental, unreplicated: Powner & Jeffery (Journal of Biophotonics 2024;17:e202300521; full text verified) gave 15 minutes of 670nm light before an oral glucose tolerance test in 30 participants. The widely quoted **27.7% is the reduction in incremental area-under-curve — the rise above baseline. The absolute AUC reduction was 7.3%, the peak-glucose reduction 12.1%**. n=30, single study, no replication. And note what the marketing on both sides misses: the light source was an LED — so this study cannot be conscripted into "LED red is degraded/harmful compared to natural red". Cui bono: quoted by panel sellers as a metabolic claim; the incremental-vs-absolute switch is the tell that a seller has not read the paper.

The dose machinery — what lets a user evaluate any device:

6. Midday daylight already delivers a therapeutic-window red/NIR dose (physics-measurement, Foundational). PHYSICS: integrating the ASTM G173 (AM1.5G) solar reference spectrum over 600–900nm gives roughly 20–35 mW/cm² in-band under clear midday sun — squarely inside the 25–50 mW/cm² window much of the photobiomodulation literature itself treats as therapeutic — full-body, free; ten minutes is on the order of 18 J/cm², at or above many tested fluences. (Band integrals are arithmetic on the standard reference spectrum, ±15%.) At sunrise/sunset the in-band irradiance collapses roughly 40–90× to ~0.5–3 mW/cm² — figures calculated from standard atmospheric attenuation physics at low solar elevation, not from measured spectral datasets, and stated as such. Golden hour is red-rich but photon-poor: its credit is circadian phase and near-zero UV, not a photobiomodulation dose. The free ticket is not cost-free either — peak free NIR and peak UV arrive on the same midday beam (see morning_sunlight_exposure and the sun-exposure balance entries).

7. Device spec sheets are metrologically unreliable (grey-literature measurement, Moderate as a caveat). PHYSICS-MEASUREMENT (independent grey literature): independent spectrometer testing of consumer devices (Outliyr, GembaRed methodology) finds advertised irradiance ranging from 33–65% under to roughly 3× over the printed spec, driven by two failure modes: quoting irradiance at 0mm (lens contact) rather than usage distance, and measuring with broadband meters that overcount. Cui bono both ways: the testers sell rankings and affiliate links; the manufacturers being tested sell the panels. The methodological point — specs at 0mm with broadband meters overstate dose — stands independent of either.

8. Mask versus panel is a dose gap, not a style choice (physics + trial-protocol comparison). PHYSICS-MEASUREMENT: typical LED mask sessions deliver on the order of 1–3 J/cm², against the committed multi-week panel protocols (10–30 sessions) the photoaging trials actually ran. A mask is plausibly the weakest form factor in its own category, marketed on the strongest sub-literature.

The hypothesis layer and the contradictions — held at their honest ceiling:

9. The "infrared deficiency" hypothesis (Experimental — firewalled as hypothesis). HYPOTHESIS PAPERS: the Zimmerman & Reiter lineage (Melatonin Research 2019; Tan et al., Biology 2023) proposes that indoor living deprives humans of a systemically necessary NIR input (stacking sub-hypotheses about mitochondrial melatonin and deep-tissue photon penetration). It is self-labelled as hypothesis, none of the stacked links has outcome evidence, and the lead author sells NIR lighting products. A mechanistic counterargument from the photobiomodulation field itself: the 700–770nm "dead zone" — an absorption trough of the proposed cytochrome-c-oxidase chromophore — sits in the middle of the band a blanket "all NIR is nourishing" claim would need to be flat across (see mitochondrial_health).

10. The biphasic dose response, as used, is unfalsifiable (methodological observation). The field's own doctrine says too little light does nothing and too much inhibits ("biphasic dose response"). In practice this explains every null result after the fact — wrong dose, wrong wavelength, wrong schedule. The observation that breaks it as commonly deployed: outdoor workers receive 100–1000× the "optimal" cumulative in-band fluence with no described overdose syndrome. Either the inhibition arm is far weaker than the framing implies, or the model as used is a get-out clause, not a prediction.

11. The IR-A photoaging contradiction (mechanistic, human in-vivo + in-vitro — verified). RODENT/IN-VITRO/HUMAN-MECHANISTIC: the Krutmann group's infrared work (Schroeder et al., Journal of Investigative Dermatology 2008, PMID 18449210; and related fibroblast studies) showed that infrared-A radiation induces matrix metalloproteinase-1 (MMP-1) — a collagen-degrading enzyme — in cultured human dermal fibroblasts and in human skin in vivo, via ERK1/2 signalling, without a compensating rise in its inhibitor TIMP-1 — a photoaging mechanism. So the same broad waveband is sold as anti-aging by the device industry and implicated in aging by the photodermatology literature. The likely resolution is dose and context (solar-intensity IR-A heat loads versus low-irradiance LED sessions), but no trial adjudicates it, and this is mechanistic evidence on both sides — not clinical wrinkle outcomes. Cui bono: the IR-A-damage literature partly underwrites "infrared protection" sunscreen additives — that side sells product too.

12. The absence, stated as absence. No human trial compares broadband solar NIR against narrowband LED at matched fluence. Every "sun versus panel" claim in either direction is extrapolation. Nobody has run the head-to-head.

13. The citation-laundering warning. HUMAN EXPERIMENTAL (misattributed in circulation): the skin-light systemic blood-pressure finding (PMID 24445737, n=24) is a UVA study — nitric-oxide release from UVA-irradiated skin — routinely quoted in red-light marketing as if red/NIR lowered blood pressure. Wrong waveband. Name the move when you see it.

Mechanism

The mainstream proposed mechanism is photon absorption by cytochrome c oxidase (complex IV of the mitochondrial electron transport chain), whose copper centres absorb in the red and near-infrared, transiently increasing electron transport, ATP output, and low-level reactive-oxygen signalling that shifts cells toward repair programmes. This is well-characterised in vitro and in animal models, and it is why the field's action spectra cluster around ~630–670nm and ~810–850nm — with the 700–770nm absorption trough between them, which is the single most useful fact for auditing device marketing: a claim that is indifferent to wavelength across that trough is not tracking its own proposed mechanism (mechanistic detail deferred to mitochondrial_health; the same chromophore logic appears in methylene_blue_cognition_and_mitochondria).

Two honest limits on the mechanism story. First, the mechanism-to-clinic gap: cytochrome-c-oxidase photochemistry is real chemistry, but real chemistry does not guarantee clinically meaningful effects at consumer doses through clothing-free skin at usage distance — the per-indication trial evidence above, not the mechanism, carries the claims. Second, the mechanism cuts both ways at higher doses: the Krutmann IR-A work shows the same broad waveband driving MMP-1 collagen-degrading signalling in human skin at solar-scale exposures. Photons in this band are a dose-dependent biological input, not a nutrient.

The hormesis frame is the right mental model and also the marketing's favourite escape hatch: a low-dose stimulus with a plausible inverted-U response (see hormesis_and_adaptive_stress) — legitimate as biology, unfalsifiable when invoked after the fact to explain away every null result (the biphasic point above).

Risks And Contraindications

• Eyes near bright emitters. The one published mask-associated retinal injury case (PMC7302677) involved a blue-emitting mask — the injury does not transfer to red/NIR wavelengths, but it establishes that face-worn consumer light devices can reach retinally significant intensities. Use supplied eye protection; do not stare into emitters; keep blue-emitting masks away from unprotected eyes entirely.
• Engineered high-irradiance devices genuinely can injure retinas. Repeated low-level red-light (RLRL) myopia devices — an adjacent, prescription-adjacent category — have a published retinal-injury case (JAMA Ophthalmology 2023) and independent measurements exceeding laser safety classification limits (Ostrin & Schill 2024), and China reclassified them as Class III medical devices. Intensity, not spectrum ideology, is the hazard variable: screens cannot injure your retina; a misdesigned engineered emitter can (full treatment in outdoor_light_and_childhood_myopia).
• Heat and high cumulative IR-A dose. The Krutmann mechanism work implicates high infrared exposures in collagen-degrading signalling — another reason "more is better" is the wrong frame for panels used at close range for long sessions.
• Photosensitising medication and active skin conditions. Standard caution: photosensitising drugs, active skin malignancy or undiagnosed lesions in the treatment field warrant medical clearance first — trials excluded these populations.
• Active eye disease. Anyone with retinal disease considering light-based eye claims belongs in ophthalmology, not in a consumer purchase funnel built on a single-lab n=20 finding.

Controversy

Nature of the controversy. Whether consumer photobiomodulation is an evidence-backed therapy or a wellness-industry dose of real-science aesthetics — fought between a device industry quoting its own sponsored trials at maximum generality and a skeptic reflex that rounds the whole category to placebo.

Position A — "it works; the science is extensive." Best evidence: the AGA meta-analysis (eleven double-blind sham-controlled RCTs, large pooled effect) is genuinely strong by consumer-wellness standards; the skin RCTs are real; the mechanism is characterised chemistry, not hand-waving. Where it overreaches: exporting hair-and-skin evidence to metabolic, cognitive, hormonal and "cellular energy" claims with no outcome trials; hiding the FDA-cleared/approved distinction; quoting panel protocols to sell masks; quoting incremental AUC as if it were absolute glucose reduction; and never once printing the midday-daylight arithmetic that competes with the product at $0.

Position B — "it's all placebo and red-tinted marketing." Best evidence: sponsorship saturation of the trial base; the narrow authorship clusters in the muscle literature; unreplicated single-lab findings driving viral claims; spec sheets that fail independent measurement; the unfalsifiable deployment of biphasic dosing. Where it overreaches: eleven double-blind sham-controlled trials in one indication is not placebo-tier, and "the sun does it for free" — while true at midday — concedes the mechanism it claims to be dismissing.

The funding/bias dimension — cui bono, both ways. Nearly every actor in this conversation sells something: device manufacturers fund the supporting trials; grey-literature "independent testers" sell rankings and affiliate links; the NIR-deficiency hypothesis's lead author sells NIR lighting; sun-maximalist influencers monetise courses and supplements while ignoring both the dose arithmetic and the UV coupling; even the IR-A-damage literature partly underwrites infrared-protection cosmetic additives. The platform sells none of it — which is what allows both halves of this entry to be stated in the same document.

Realised Position: per-indication tiers, stated plainly: hair Strong-with-caveats, skin Moderate on panel protocols, muscle Moderate at low certainty, eyes Emerging, glucose Experimental, systemic "light nutrition" claims firewalled as hypothesis. The consumer headline is the cost inversion — everything with Strong evidence in the light family is free, and money flows toward the weakest tier, with the caveated hair helmets the lone exception. Daylight, mid-morning to mid-afternoon, is the free red/NIR source; sunrise keeps its circadian credit and loses its dose claim. A device purchase is defensible for hair, arguable for committed skin protocols, and premature for everything else.

Cross-Pillar Connections

• light_environment_and_health — the family hub; carries the tier map and the cost-inversion doctrine this entry instantiates for the device aisle.
• artificial_light_vs_sunlight — the one-stop on light quality; owns the spectral-harm firewall and the daylight-as-free-red/NIR correction this entry's dose section details.
• morning_sunlight_exposure — sunrise's genuine credit (circadian phase, near-zero UV) lives there; this entry supplies the reason it is not a red-light dose.
• outdoor_light_and_childhood_myopia — owns the RLRL device story in full: the efficacy trials, the retinal-injury case, and the intensity-not-spectrum hazard carve-out this entry points to.
• mitochondrial_health — the cytochrome-c-oxidase mechanism, the 700–770nm dead zone, and the mechanism-to-clinic gap are grounded there.
• methylene_blue_cognition_and_mitochondria — the sibling "boost the electron transport chain" consumer claim; same chromophore logic, same tier discipline.
• looksmaxxing_health_first_evidence — hair and skin are appearance indications; that entry's health-first doctrine frames when a cosmetic device purchase is sane versus signal-chasing.
• hormesis_and_adaptive_stress — the inverted-U dose model this category invokes; that entry carries the honest version, this one documents its unfalsifiable deployment.

What would change our mind

Industry bias note

Sources (14)

• Liu et al. (Liu, Liu, Chen & Chin), 2019, comparative effectiveness of low-level laser therapy for adult androgenic alopecia, Lasers in Medical Science (pubmed.ncbi.nlm.nih.gov/30706177↗/" target="_blank" rel="noopener">PMID 30706177↗) (constituent trials largely manufacturer-sponsored) — 11 double-blind sham-controlled RCTs; hair density SMD 1.316 vs sham; absolute gains ~+14–19 hairs/cm².
• Wunsch & Matuschka, 2014, controlled trial of red and near-infrared light for skin rejuvenation, Photomedicine and Laser Surgery (randomised, four arms including a placebo-equivalent control; twice-weekly sessions for 15 weeks) — n=136; improved complexion and collagen-density measures on multi-week panel-class protocols.↗
• Mota et al., 2023, periocular photobiomodulation RCT (pubmed.ncbi.nlm.nih.gov/36780572↗/" target="_blank" rel="noopener">PMID 36780572↗) (split-face, red vs amber — both arms active, no sham arm; 10 sessions over four weeks) + systematic review PMC11049838 (reviews decline to pool) — real controlled photoaging signal; heterogeneity blocks meta-analysis.
• DOMS/pre-exercise photobiomodulation meta-analyses (PMC12286287; a 2024 meta pooling 34 RCTs) (narrow, partly device-linked authorship cluster) — moderate effects, low certainty.↗
• Shinhmar et al., 2021, weeklong improved colour contrast sensitivity after single 670nm exposures, Scientific Reports 11:22872; and Shinhmar et al., 2020, Journals of Gerontology A (pubmed.ncbi.nlm.nih.gov/32596723↗/" target="_blank" rel="noopener">PMID 32596723↗) (academic lab; no direct commercial stake identified; heavily merchandised by third parties) — n=20, single 3-min morning exposure, ~17% mean improvement, morning-only; single lab, unreplicated.
• Powner & Jeffery, 2024, 670nm light and glucose tolerance, Journal of Biophotonics 17:e202300521 (academic; full text verified) — n=30; incremental AUC −27.7%, absolute AUC −7.3%, peak −12.1%; LED source; unreplicated.↗
• ASTM G173 / AM1.5G solar reference spectrum (physics standard; no stake) — midday ~20–35 mW/cm² in 600–900nm; sunrise/sunset in-band ~0.5–3 mW/cm² by standard atmospheric-attenuation calculation (calculated, not measured).↗
• Independent device spectrometer testing (Outliyr / GembaRed methodology) (testers sell rankings and affiliate links) — advertised irradiance 33–65% under to ~3× over; 0mm and broadband-meter failure modes.↗
• Blue-emitting mask retinopathy case report (PMC7302677) (case report, n=1) — the one published mask retinal injury; blue wavelengths, does not transfer to red/NIR.↗
• Liu et al., 2023, RLRL retinal injury case, JAMA Ophthalmology; Ostrin & Schill, 2024, RLRL device safety measurements, Ophthalmic and Physiological Optics 44:241 (independent academic safety work) — engineered high-irradiance devices exceed laser safety classification limits; injury documented.↗
• Schroeder et al., 2008, infrared radiation-induced matrix metalloproteinase in human skin, Journal of Investigative Dermatology (pubmed.ncbi.nlm.nih.gov/18449210↗/" target="_blank" rel="noopener">PMID 18449210↗), and related Krutmann-group fibroblast studies (IR-damage literature partly underwrites infrared-protection cosmetic additives) — IR-A induces MMP-1 in human dermal fibroblasts and human skin in vivo via ERK1/2, without TIMP-1 compensation.
• Zimmerman & Reiter, 2019, Melatonin Research; Tan et al., 2023, Biology (lead author sells NIR lighting products) — the NIR-deficiency hypothesis lineage; self-labelled hypothesis, no outcome evidence.↗
• UVA skin irradiation and blood pressure study (pubmed.ncbi.nlm.nih.gov/24445737↗/" target="_blank" rel="noopener">PMID 24445737↗) (academic; n=24) — the systemic light-skin finding routinely laundered into red-light marketing; it is a UVA result.
• Absence: no human trial comparing broadband solar NIR vs narrowband LED at matched fluence — stated as absence.↗

Related entries