Extracorporeal shockwave therapy for the knee

A shockwave is one pressure pulse with a very fast rise time, generated outside the body and focused into tissue. The same physics breaks kidney stones; aimed at a knee at a small fraction of that energy, it has become one of the more widely offered clinic treatments for osteoarthritis, and one of the better supported on symptoms.

Dose is stated two ways: energy flux density in millijoules per square millimetre, and a pulse count. A knee protocol usually runs 800 to 2,000 pulses at 0.05 to 0.25 mJ/mm², weekly or twice weekly for three or four weeks.

What the pooled trials show

The umbrella review sits on top of eight meta-analyses, each rated between 8 and 11 on AMSTAR 2, searched to September 2023. Against sham, shockwave improved the WOMAC by 2.94 points, the visual analogue pain scale by 2.0 points and range of motion by 17.55 degrees, with the Lequesne index moving by 2.85 points.

Two things about that summary are worth carrying. The pain and range-of-motion estimates carried no heterogeneity at all, which is unusual in this field. And the review's own limitations say that key primary studies appear in more than one of the pooled meta-analyses, so the trials underneath are counted repeatedly, and that energy levels could not be separated.

In the Bayesian network that ranks nine interventions across 32 trials, shockwave was one of only two that beat resistance and range-of-motion exercise on pain — second to low-level light on both pain scales — and first of six on the six-minute walk. Those comparisons between devices are the network's indirect estimates. No trial in it put shockwave against light directly.

The picture is not uniform. In the review that isolates what a device adds to exercise, shockwave did not separate from a placebo device with exercise, or from exercise alone, on either pain or function, on very low certainty from two trials.

The sham-controlled pilot

Eighteen people with mild knee osteoarthritis, randomized to active or sham at 0.05 mJ/mm², a thousand pulses weekly for three weeks. Both groups improved on pain, the WOMAC and the Lequesne index, so the symptom gain was not specific to the shockwave.

What was specific: suprapatellar effusion height fell in the treated group and differed from control at one month. Knee flexion range and Doppler activity rose immediately after the last session and the difference was gone by the follow-up.

It is a small trial at an energy five times below the usual protocol, and it is the clearest illustration in this shelf of why an uncontrolled before-and-after improvement means little here.

What happened when someone imaged the cartilage

Two studies looked.

The larger is not randomized: 82 patients treated with radial shockwave at 0.25 mJ/mm² twice weekly for four weeks, compared with 104 patients from the previous year who took oral NSAIDs. Pain and function improved from baseline and peaked eight weeks after treatment. Against the NSAID group there was no significant difference in pain or WOMAC at four or eight weeks.

At 24 weeks, cartilage lesion score and T2 mapping across the patellar, trochlear, medial and lateral femoral, and medial and lateral tibial regions of 76 injured knees were unchanged from before treatment. The authors say plainly that they set out expecting the cartilage regeneration the animal work reports, and did not find it.

The smaller is the sham-controlled pilot: cartilage thickness at the medial trochlea, trochlear notch and lateral trochlea did not change over time and did not differ between groups.

A 2026 mechanistic review reaches the same place from the other direction, stating that evidence for structural cartilage regeneration comes mainly from animal models, small biomarker studies and heterogeneous protocols, and that long-term human imaging evidence is scarce.

The mechanism, and where it points

Two accounts run through this literature and they are not the same claim.

The analgesic account is neural: shockwave is proposed to lower sensory nerve excitability and substance P and CGRP signalling, damping peripheral sensitization and synovial inflammation. That is the account the human evidence best fits.

The structural account is indirect and specific — shockwave acts on subchondral bone, and cartilage benefits because the bone beneath it remodels. The sharpest version of it is a rat experiment that aimed at bone in one group and at the cartilage surface in another. Aiming at the medial tibial subchondral bone won on histology, micro-CT, cartilage grading and synovial grading, and in the cartilage of those animals TGF-β1 rose while DMP-1, MMP-13 and ADAMTS-5 fell.

Two companion studies from the same group found the same ordering: treating the medial femoral and medial tibial condyles together gave the best modified Mankin score, the largest sectional cartilage area and the least proteoglycan loss, and in a seven-group comparison the medial tibia beat every other aiming point on OARSI score, cartilage defect size and bone mineral density.

That is a real result and it comes with a real caveat: three studies, one laboratory in Kaohsiung, one device, one surgical model, eight animals a group, and one author on the advisory committee of the company that makes the device.

Safety

No serious adverse event appears in the human studies here. All eighteen patients in the sham-controlled pilot completed treatment without complication, and the cohort study reports none. Transient local discomfort during treatment is described as usual and tolerable. In rats, the standard dose damaged neither meniscal cartilage nor joint tissue against sham.

What would change this

Three specific studies, none of which exists. A trial that randomizes energy level — the parameter the umbrella review could not analyse and the mechanism reviews say is undetermined. A human trial that aims at subchondral bone the way the animal work says to aim, and reports where it aimed. And a two-year imaging trial at a protocol already shown to work on symptoms, which is the only way the structural question gets answered rather than deferred.