Red Light Therapy: What the Research Actually Shows About Photobiomodulation.

In 1967, a Hungarian physician named Endre Mester aimed a ruby laser at the shaved backs of mice, expecting to induce tumours.^[1] The tumours never came. What came instead was something no one was looking for: the mice regrew their fur faster than the unirradiated controls, and their wounds healed at an accelerated rate. Mester had stumbled onto a therapeutic effect he could not explain. A low-power light beam, too weak to cut or cauterise, was somehow accelerating biological repair. He spent the remaining seventeen years of his career pursuing a mechanism that eluded him.

That accidental discovery launched a field that has spent six decades caught between two competing realities. On one side: a growing body of controlled trials demonstrating measurable effects on pain, muscle recovery, skin remodelling, and inflammation. On the other: a consumer market flooded with unregulated devices, implausible marketing claims, and a reputation problem so severe that many clinicians still dismiss red light therapy as pseudoscience without examining the trial data.

The gap between what the research shows and what the market sells is the central tension of photobiomodulation, the formal name for the therapeutic use of red and near-infrared light at non-thermal power densities.^[3][8] Understanding that gap requires looking at what 204 randomised controlled trials actually measured, what they found, and why the results have been so inconsistent across studies that used nominally the same intervention.

The scale of the evidence base has shifted substantially in the past decade. Son et al.'s 2025 umbrella review, the most comprehensive assessment of PBM to date, synthesised fifteen meta-analyses drawing on 204 RCTs, more than 9,000 participants, across 32 countries and six continents.^[4] The review found moderate-certainty benefit for five conditions: fibromyalgia fatigue, androgenetic alopecia, knee osteoarthritis disability, burning mouth syndrome pain, and age-related cognitive impairment. No outcome reached high certainty. That distinction matters, and this article will not pretend otherwise.

What the umbrella review also revealed is that 57% of the 35 measured health endpoints showed only low certainty, and another 26% showed very low certainty.^[4] The biological signal is real. The measurement precision is not yet where it needs to be. And the reason, as we will see, has far more to do with how light was delivered than whether it worked.

Even the language tells part of the story. The field has cycled through names: "low-level laser therapy," "cold laser," "low-power light therapy," each carrying different clinical and commercial baggage. The 2015 consensus adoption of "photobiomodulation" was itself an attempt to separate the biology from the branding.^[3] The Cochrane review on seasonal affective disorder uses "light therapy" in yet another sense, referring to bright white light and dawn simulation, not red/NIR. It illustrates how the same word can mean different interventions in different clinical contexts.^[6]

The story of photobiomodulation begins with a single enzyme. In 1999, Tiina Karu identified cytochrome c oxidase, the terminal enzyme in the mitochondrial electron transport chain, as the primary intracellular photoacceptor for red and near-infrared light.^[7] That identification gave the field a mechanism it had lacked for thirty years. When photons in the 630–1100 nm wavelength window strike cytochrome c oxidase, they displace nitric oxide molecules that have bound to the enzyme's copper and heme centres, inhibiting electron transport.^[9] Remove the nitric oxide, and electron flow resumes. The mitochondrion starts producing adenosine triphosphate again. That is the primary event.

The mechanism is photochemical, not thermal. The power density involved, typically 1–100 mW/cm², produces negligible tissue heating.^[16] What it produces instead is a molecular signal. The displaced nitric oxide diffuses into the surrounding tissue, triggering vasodilation. A small burst of reactive oxygen species activates protective signalling pathways. And the restored electron transport chain begins generating ATP at an increased rate.^[10]

This is where the mechanism becomes important for understanding the field's inconsistent results. The effective fluence window, the dose of light energy per unit area, is narrow: 0.5–10 J/cm² for most applications.^[16] Below that range, insufficient photons reach the chromophore. Above it, the reactive oxygen species burst shifts from protective to damaging. Huang et al. demonstrated this biphasic pattern directly: in mouse traumatic brain injury models, 36 J/cm² was optimal, while 360 J/cm² eliminated the benefit entirely.^[14][15]

The downstream cascade is where the biology gets interesting. Once cytochrome c oxidase resumes normal function, three parallel signalling events unfold. The ROS burst, small enough to be protective rather than destructive, activates NF-κB, a transcription factor that in normal cells upregulates genes for proliferation, migration, and anti-apoptotic signalling.^[10][13] The displaced nitric oxide triggers endothelial nitric oxide synthase (eNOS), amplifying vasodilation beyond the initial photodissociation event.^[12] And calcium ion channels open, feeding into Akt/GSK3β signalling pathways that regulate cell survival and growth.^[10]

Hamblin has described photobiomodulation as an "exercise mimetic," a framing that captures something important about the dose-response logic.^[9] Physical exercise also produces a controlled ROS burst that activates protective gene expression. Too little exercise produces no adaptation. Too much produces damage. The parallel is not metaphorical; it is mechanistic. Both interventions exploit the same biphasic signalling window.

The wavelength specificity reinforces this. Zein et al.'s parameter review established that red light at 630–670 nm is effective for superficial targets, near-infrared at 780–940 nm penetrates to deeper tissue, and the 700–770 nm gap between them is largely ineffective: a dead zone where neither cytochrome c oxidase absorption nor tissue penetration is sufficient.^[16] Consumer devices that operate in this gap produce no measurable effect. That is not mechanism failure; it is wavelength mismatch.