Lasers and LEDs for knee osteoarthritis

Shine red or near-infrared light on a knee and some of it is absorbed by the mitochondria inside the cells beneath. That is the premise of photobiomodulation, and it has been tested on knee osteoarthritis more than on any other joint problem — in lasers held against the skin, in laser needles at acupuncture points, in high-powered clinic devices, and in the LED panels sold for use at home.

The trials are numerous and they do not all agree. What makes them readable is that the disagreement is not random: it tracks the device and the dose. A 2026 review gathered fifty-nine human studies, mostly randomized trials and systematic reviews, and deliberately declined to pool them, sorting them into six protocol classes instead. That decision is the reason this entry can be organised the way it is, and it is also the review's main limitation.

Everything below is a symptom endpoint unless it says otherwise: pain on a visual analogue or WOMAC scale, function, range of motion, measured over three to eight weeks.

Low-level laser against placebo

Ten randomized trials and three systematic reviews, at 633 to 905 nm and most often 830 or 904. Courses ran eight to twenty sessions across three to eight weeks, delivering 5.76 to 27 joules per session.

The result is genuinely mixed. Many trials report pain reduction against sham; some report none. The earliest of them dates to 1992, so this is not a young literature, and its persistence in disagreement after three decades is what the review is trying to explain when it points at unstandardised parameters.

Laser acupuncture

Nine randomized trials and one systematic review, delivering light to acupuncture points rather than across the joint surface, at 650 to 10,600 nm — often 780 to 830 nm combined with a carbon-dioxide laser.

This is the one protocol class where the review reports no internal disagreement: consistent short-term pain reduction and functional improvement. Worth knowing that "short-term" is doing real work in that sentence, and that only two studies in the wider preclinical literature used the carbon-dioxide laser at all, which limits how much can be said about that half of the technique.

Light added to an exercise programme

Eleven randomized trials and two systematic reviews, at 808 to 904 nm and also 850 and 880, over ten to twenty-four sessions. Generally superior to exercise alone, though some of the systematic reviews disagree, and one group that reviewed seven trials of the combination concluded the question remains controversial.

A detail worth carrying: in the animal literature the combination does nothing. The systematic review of thirty-three cell and animal studies found that adding exercise, or topical anti-inflammatories, to light therapy gave no advantage over light alone — while adding intra-articular stem cells, or chondroitin with glucosamine sulfate, did. The species disagree about which partners help, which is a reason to hold the human combination result loosely rather than to discard it.

High-intensity laser

Twelve randomized trials and four systematic reviews, usually at 1,064 nm, generating output above 500 milliwatts and delivering 79.2 to 3,000 joules per session. It consistently reduces pain and is often superior to low-level laser. One trial found seven sessions at 1,064 nm and 12 watts effective for pain and function against placebo.

This is a clinic modality and worth naming as one. It works by a partly different route — deliberate surface hyperthermia driving oxidative reactions and ATP production, rather than the purely photochemical effect of the low-power devices — and it is delivered by devices in a regulatory class that restricts who operates them.

LED devices

Four randomized trials, at 640 to 905 nm, two to three sessions a week for two to six weeks. The findings are mixed: one study found no benefit from LED alone, while others found it comparable to low-level laser when combined with exercise.

So the format that is cheapest, wearable and easiest to own has the smallest clinical literature of the six classes here. Why that is, and what it implies for a panel bought online, is the subject of a separate entry on dose and depth.

Light against the other physical modalities

Six randomized trials put low-level laser head to head with other physical treatments, at 810 to 980 nm. It came out similar to neuromuscular electrical stimulation, and less effective than ozone or pulsed electromagnetic field.

That comparison is rarely quoted and belongs here. The review also notes that no research compares low-level laser against therapeutic ultrasound, extracorporeal shockwave or short-wave therapy at all, so the modality's standing among its neighbours is only partly mapped.

What the trials measured besides pain

Four trials in this literature looked past symptoms at the tissue, and they disagree.

At 905 nm, low-level laser gave short-term improvement in pain, joint space width, and a panel of turnover markers — CTX-II, MMP-3, MMP-8 and MMP-13. Ten sessions of high-intensity laser with exercise beat placebo laser with exercise on pain, WOMAC, range of motion and femur cartilage thickness. A second high-intensity laser trial, combining it with glucosamine sulfate and exercise, reduced pain, WOMAC subscales and synovial thickness — and found no significant difference in medial or lateral femoral cartilage thickness across its three protocols. A fourth trial raised the anti-inflammatory cytokine IL-10 and left CTX-II unmoved.

Small trials, structural measures carried as secondary outcomes, results pointing in different directions. The review's own conclusion is the careful one: definitive clinical evidence of joint space widening or structural reversal is limited, and cartilage regeneration in humans remains emerging and partly speculative because it is extrapolated from preclinical histology.

How the trials were run and appraised

The included randomized trials were scored on the PEDro scale, a ten-point measure of methodological quality. Scores ran from 5 to 10 with a mean of 6.9 — medium to high — and the review reports a gradient worth knowing: trials scoring 7 or above tended to show more consistent short-term benefit where parameters and exercise combinations were appropriate. Better-conducted trials found the effect more reliably, which is the direction a real effect points.

Against that: the synthesis is narrative. Nothing is pooled, so no effect size or confidence interval comes out of it, and "mixed" here means a count of trials rather than a statistical result. The clinical search used one database and stopped at the end of 2023. A separate comprehensive review of light therapy across both rheumatoid arthritis and osteoarthritis reaches a compatible verdict from another direction — cataloguing how trials were evaluated, noting that some rest on questionnaire responses, and stating plainly that no optimal parameter set has been agreed for wavelength, dose, power density, irradiation time or treatment site.

Safety

The reported adverse effects are mild and local: skin irritation, itching and redness, described as not harmful. Treatment is non-invasive and the low-power classes produce no thermal sensation.

The real hazard sits with the high-power devices rather than the therapy. Medical lasers fall in regulatory classes III and IV, are potentially hazardous if mishandled, and carry restrictions on use — which is the practical reason high-intensity laser is something done to you in a clinic rather than something bought.

What would change the tier

A pooled meta-analysis with a formal risk-of-bias assessment, on a standardised protocol, would settle whether the mixed results are heterogeneity or noise. The PEDro gradient already hints at the answer.

For a claim about the joint rather than the symptom, the study to run is one powered on an imaging or turnover endpoint from the outset, rather than carrying it as a secondary measure — long enough to see structure move. Four trials have looked; none was built for the question.