Most treatments are described by what they are. Photobiomodulation is better described by how it is delivered. Two clinics can both offer "laser for tennis elbow," bill the same code, and be doing measurably different things — different enough that in the pooled trial data, one approach produced a clear benefit and the other produced nothing at all.
That sounds like the kind of claim a clinic makes to distinguish itself. It is not. It comes from a single open-access meta-analysis, and the numbers are unusually clean.
What 13 trials showed
Bjordal and colleagues, BMC Musculoskeletal Disorders 2008 pooled 13 randomized trials covering 730 patients with lateral elbow tendinopathy — tennis elbow. Alongside the usual meta-analysis, they did something less common: they assessed the procedures themselves, recording what wavelength each trial used and where on the arm the light was aimed.
Pooled across everything, pain relief was WMD 10.2 mm (95% CI 3.0 to 17.5), p = 0.005 on a 100 mm visual analog scale. Statistically real. Clinically, not much. On its own, that number would justify a shrug.
Then they split the trials by wavelength and target.
904 nm, on the tendon
Trials using 904 nm applied directly to the tendon: WMD 17.2 mm (95% CI 8.5 to 25.9), p = 0.0001.
Trials using 820 nm, 830 nm or 1064 nm: no significant effect.
Trials that irradiated acupuncture points instead of the tendon: negative results.
Read those three lines together and the shrug becomes something more interesting. The pooled 10.2 mm was not a description of a modestly useful treatment. It was an average of a treatment that worked in the trials that did one specific thing and did not work in the trials that did something else. Averaging them together produced a number that describes neither group.
Function moved in the same direction. Pain-free grip strength — how hard you can squeeze before the elbow stops you, which is closer to a daily-life measure than a pain line — favored laser over placebo at SMD 0.66 (95% CI 0.42 to 0.90).
Two variables account for most of the story here: the wavelength of the light, and whether it was aimed at the injured structure. Neither is exotic. Both are things a clinic either knows about its own protocol or does not.
The finding that cuts the other way
The same paper reports something clinics quoting it usually leave out. Using Egger’s test, the authors detected a considerable degree of publication bias in this literature.
Publication bias means the published trials are not a representative sample of the trials that were run. Studies that found nothing are systematically less likely to appear in journals, which means any pooled estimate built from published work can be distorted. It does not tell you the 17.2 mm figure is wrong. It tells you the figure rests on a foundation that is known to be uneven, and that a reader should hold it more loosely than the tight confidence interval suggests.
That is one reason we describe the overall certainty here as low rather than moderate, and why the next two sections matter as much as the one above.
The same pattern, in the shoulder
If dose and delivery really are what separate a result from no result, the effect should show up in other tendons. It does.
Haslerud and colleagues, Physiotherapy Research International 2015 reviewed 17 randomized trials of laser therapy for shoulder tendinopathy. Laser used alone gave WMD 20.41 mm (95% CI 12.38 to 28.44); laser combined with exercise gave 16.00 mm (95% CI 11.88 to 20.12).
Their conclusion about the trials that failed is the sentence worth carrying around: "Trials performed with inadequate laser doses were ineffective across all outcome measures."
Not less effective. Not effective for pain but not function. Ineffective across all outcome measures. A different research group, a different joint, the same conclusion Bjordal reached about wavelength and target — get the parameters wrong and you have not delivered a weaker version of the treatment. You have delivered something that does not register.
Both groups’ recommended parameters are published in the WALT dosage tables, which specify wavelength, energy per point and number of points separately for each wavelength range. The 780–860 nm table and the 904 nm table are constructed differently and should not be mixed.
Where tendinopathy sits in the wider evidence
It would be easy to stop here and conclude that light aimed correctly at tendons is a settled matter. The broadest review available says otherwise.
Son and colleagues, Systematic Reviews 2025 assembled an umbrella review of 15 meta-analyses covering 204 randomized trials, more than 9,000 participants and 35 outcomes across 15 conditions. Their headline finding: "No outcome was supported by high certainty of evidence." Not one, anywhere in the field.
And specifically on tendons: Achilles tendinopathy and carpal tunnel syndrome showed no significant effect.
That deserves to be stated without softening. The tendinopathy evidence overall is low certainty. Lateral elbow is the site where the wavelength-and-target signal is clearest, and even there the underlying literature carries detectable publication bias. Somebody who tells you photobiomodulation reliably treats tendon problems is describing a body of evidence that does not exist.
Which brings us to the part that actually does the work
For tennis elbow, and for Achilles tendinopathy in particular, the treatment is loading exercise.
Progressive loading — controlled, gradually increasing demand on the tendon, usually eccentric or heavy slow resistance work, done consistently over months — is the intervention with the strongest support and the one that changes how the tendon tolerates use. It is unglamorous, it is often uncomfortable early on, and it requires the patient to keep doing it after the appointments stop. There is no version of tendon rehabilitation that skips it.
Light, at best, sits next to it. The clearest illustration is Stergioulas and colleagues, American Journal of Sports Medicine 2008, a randomized trial of 52 recreational athletes with chronic Achilles tendinopathy. Every participant did eccentric exercise. Half also received laser, half received sham. Pain during activity, laser versus sham: 53.6 vs 71.5 mm at 4 weeks (P = .0003), 37.3 vs 62.8 at 8 weeks (P = .0002), 33.0 vs 53.0 at 12 weeks (P = .007).
Note what that trial was not. It was not laser against exercise. It was exercise plus laser against exercise plus sham. The authors framed the result as accelerating recovery on top of eccentric exercise, not as a treatment in its own right — and the umbrella review above still classifies Achilles tendinopathy as showing no significant effect once the wider literature is pooled. Both of those things are true at once, and holding them together is the honest position.
What to ask
Since the delivery is the treatment, these are the questions that separate one clinic’s protocol from another’s:
- What wavelength is the device?
- Is the applicator placed on the tendon itself, and how is that location identified?
- How many joules per point, how many points, how many sessions?
- Do those parameters correspond to the WALT recommendations for that specific wavelength?
- What am I doing for loading exercise, who is prescribing and progressing it, and how is it being tracked?
If the answer to the last question is vague, the first four do not matter much.
The takeaway
Technique specificity in this modality is real and it is measurable. In 730 patients, 904 nm aimed at the tendon produced 17.2 mm of pain relief; three other wavelengths produced no significant effect; and aiming at acupuncture points instead of the tendon produced negative results. That is not a marketing distinction. It is a subgroup analysis in an open-access paper anyone can read.
This is not a modality where "some laser" is equivalent to the right wavelength, on the right structure, at the right dose. And even done exactly right, it remains an addition to loading exercise — never a substitute for it.
This article is general education, not medical advice, and it is not a substitute for evaluation by a qualified clinician. It does not establish a physician–patient relationship. Individual results vary, and the studies described report group averages rather than what any one person should expect. Treatments discussed may not be FDA-cleared or approved for the conditions mentioned; where that is the case it is stated in the text. Talk to your own physician before starting, stopping or changing any treatment.