Get urgent care for a knee that is suddenly hot, red, or very swollen, a fever, not being able to put weight on your leg, or calf swelling or shortness of breath.

Key facts

  • Sham-controlled trials are mixed: one small trial found less pain after GAE than after a sham procedure, and two larger trials found no clear difference5,6,7.
  • Studies without a comparison group report less pain after GAE, but they cannot separate the effect of GAE from the placebo effect8.
  • In one early single-center, uncontrolled study, about 80% of patients still had meaningful relief 3 years after GAE9.
  • Some people may still need a knee replacement later. In one study, 9 of 46 people later had one10.
  • Evidence status as of October 2026: more randomized trials are still underway2,3,4.

Two kinds of studies

Most GAE studies are open-label. Everyone gets GAE and knows it, and researchers track pain scores over time. A 2025 review that pooled these studies found that pain scores dropped by about 30 points on a 100-point scale on average8.

Open-label studies show whether people felt better, but not why. Knee arthritis pain often goes up and down, and expecting a treatment to help can ease pain on its own. This is called the placebo effect. In a sham-controlled trial, some people get GAE and others get a sham (fake) procedure, and they are not told which one they got. If the GAE group does better, the difference is more likely to come from GAE itself.

Questions to ask

  • Which studies are you basing your advice on?
  • Did those studies compare GAE with a sham procedure or another treatment?
  • How do you measure pain before and after GAE in your own patients?

Sham-controlled trial results

Three sham-controlled trials have reported results so far.

Sham-controlled trials of GAE for knee arthritis pain
TrialWho took partWhat it found
Bagla et al., 2022 21 people with mild to moderate arthritis, in a single-blind trial (patients were not told which procedure they got) At 1 month, pain improved more after GAE, by about 50 points more on a 100-point scale. Everyone in the sham group then crossed over to GAE5.
Landers et al., 2023 59 people with grade 2 arthritis on X-ray, in a triple-blind trial No significant difference at 12 months. Knee pain scores improved 41% after GAE and 29% after the sham procedure6.
van Zadelhoff et al., 2024 and 2026 58 people, in a double-blind trial No significant difference at 4 months or at 12 months. MRI showed no change in inflammation of the joint lining7,11.

Put together, sham-controlled trials are mixed. A 2025 review of the three trials found a short-term pain benefit in one trial, no significant gain in knee scores overall, and no serious side effects1. Another 2025 review said the benefit over placebo "remains uncertain"8. The authors of the 12-month trial wrote that their results "do not support the clinical implementation of GAE"11.

Questions to ask

  • How do these trial results apply to my knee?
  • Was the arthritis in these trials similar to mine?
  • What would make you expect more benefit for me than the trials found?

Uncontrolled and long-term studies

In one early single-center, uncontrolled study, about 80% of patients still had meaningful relief 3 years after GAE9.

In the GENESIS study of 46 people, average pain fell from about 59 to about 38 out of 100 after 2 years. Nine of the 46 later had a knee replacement, and their surgeons reported no added difficulty during the operation10.

In a 2026 study of 194 people treated with particles that dissolve over time, typical pain scores fell from 7 to 3 out of 10 after 12 months. About 1 in 5 people did not have a 12-month check12.

None of these studies had a comparison group, so they cannot show how much of the relief came from GAE itself.

How long does relief last?

Long-term data are limited. The longest follow-up comes from uncontrolled studies: relief lasted 3 years for most people in one early study9, and pain scores stayed lower at 2 years in another10. The sham-controlled trials followed people for up to 12 months6,11.

Some people may still need a knee replacement later. If pain comes back, one expert review says a repeat GAE may be considered after about 12 months, but repeat GAE has not been well studied13.

Questions to ask

  • How long do you expect relief to last for someone like me?
  • What happens if my pain comes back?

Why success rate numbers vary

A 2025 review of 23 studies found success rates from 30% to 100%14. Numbers vary because studies define success in different ways, follow people for different lengths of time, and include people with different stages of arthritis. Pain tended to improve more in mild to moderate arthritis than in severe arthritis13.

When you see a success rate, ask how success was defined, how long people were followed, and whether there was a comparison group.

Online reviews and stories

Personal stories and online reviews cannot tell you how likely GAE is to help you. People who felt better are more likely to post, and a story cannot separate GAE from the placebo effect or from pain that would have eased anyway. This page does not use testimonials. Ask your doctor for results from studies and from their own patients.

Trials still underway

Randomized trials comparing GAE with sham procedures or steroid injections are still underway (for example MOTION, GENESIS II, RESORB)2,3,4. MOTION and RESORB compare GAE with a steroid injection, and GENESIS II compares GAE with a sham procedure. You can search registered GAE trials on ClinicalTrials.gov15. Read GAE clinical trials for who can take part and how to find a trial site.

What the SIR position statement says

The Society of Interventional Radiology (SIR) published a position statement on GAE in 2026. It describes GAE for people whose pain has not improved with other care and who cannot have, do not want, or want to put off a knee replacement. It calls GAE "an established treatment option" and not an experimental one16.

SIR also says broader use "will be informed by the continued generation of high-quality clinical evidence, including randomized studies," and that ongoing randomized trials aim to measure possible placebo effects16.

Others read the evidence differently. NICE, which reviews treatments in the UK, says GAE should only be used in research17. Some commercial insurers call it investigational18,19. Read about GAE insurance and Medicare coverage.

Common questions

Does GAE really work?

Sham-controlled trials are mixed. One small trial found less pain after GAE than after a sham procedure at 1 month, and two larger trials found no clear difference at 4 to 12 months5,6,7,11. Studies without a comparison group report less pain, but cannot rule out a placebo effect.

What is the success rate of GAE?

Reported success rates range widely, from 30% to 100% in one review of 23 studies, because studies define success in different ways14. Ask how success was defined and whether there was a comparison group.

How long does GAE last?

Long-term data are limited. In one early single-center, uncontrolled study, about 80% of patients still had meaningful relief 3 years after GAE9. Some people may still need a knee replacement later.

Is GAE better than a placebo?

It is not yet clear. One small trial favored GAE, and two larger trials found no significant difference from a sham procedure1. More randomized trials are still underway2,3,4.

Why do some experts call GAE established and others investigational?

SIR calls GAE an established treatment option16. NICE in the UK says it should be used only in research17, and some commercial insurers call it investigational18,19. They weigh the same mixed trial results differently.

Bring questions to your visit

A printable list of questions about GAE, with background notes from the sources. Talk with your doctor about these questions.

Questions to ask before GAE

Talk with your doctor before you decide. This information does not replace personal medical advice. Read the medical disclaimer.

Sources

  1. Milhem et al., systematic review of sham-controlled trials of GAE, Journal of Orthopaedics 2025 (PMID 40766107) ↩︎
  2. ClinicalTrials.gov, MOTION randomized trial of GAE versus steroid injection (NCT05818150) ↩︎
  3. ClinicalTrials.gov, GENESIS II sham-controlled trial of GAE (NCT05423587) ↩︎
  4. ClinicalTrials.gov, RESORB randomized trial of GAE versus steroid injection (NCT06872567) ↩︎
  5. Bagla et al., multicenter sham-controlled trial of GAE, Journal of Vascular and Interventional Radiology 2022 (PMID 34610422) ↩︎
  6. Landers et al., triple-blind sham-controlled trial of GAE, Bone & Joint Open 2023 (PMID 37051829) ↩︎
  7. van Zadelhoff et al., double-blind sham-controlled trial of GAE, 4-month results, BMJ Open 2024 (PMID 39353688) ↩︎
  8. Abussa and Jeremic, systematic review and meta-analysis of GAE, Academic Radiology 2025 (PMID 40425436) ↩︎
  9. Okuno et al., midterm results after embolization for knee osteoarthritis, Journal of Vascular and Interventional Radiology 2017 (PMID 28365171) ↩︎
  10. Little et al., GENESIS long-term results, CardioVascular and Interventional Radiology 2024 (PMID 38819473) ↩︎
  11. van Zadelhoff et al., the same trial at 12 months, European Radiology 2026 (PMID 42009866) ↩︎
  12. Fleckenstein et al., GAE with resorbable microspheres in 194 patients, Radiology 2026 (PMID 42301014) ↩︎
  13. Brown et al., genicular artery embolization: a technical review, Journal of Clinical Medicine 2025 (PMID 40142914) ↩︎
  14. Ashraf et al., systematic review of GAE safety and results, 2025 (PMID 40219609) ↩︎
  15. ClinicalTrials.gov, registered trials of genicular artery embolization for knee osteoarthritis ↩︎
  16. Ahmed et al., Society of Interventional Radiology position statement on genicular artery embolization, Journal of Vascular and Interventional Radiology 2026 (PMID 42118083) ↩︎
  17. NICE (UK), genicular artery embolisation for pain from knee osteoarthritis, HTG595 (formerly IPG708), 2021 ↩︎
  18. Blue Cross and Blue Shield of Texas, medical policy SUR705.048, effective July 15, 2025 ↩︎
  19. Molina Clinical Policy No. 410, genicular artery embolization of the knee for osteoarthritis ↩︎