Online Tai Chi for Knee Osteoarthritis: What the Trial Shows

A free tai chi website helped knee arthritis symptoms. That does not make it magic.

Knee osteoarthritis is common, painful, and boring in the way many huge public-health problems are boring. It affects an estimated 654 million people worldwide aged 40 and older, yet many patients still do not get structured exercise support even though exercise is standard first-line care. The interesting question is not whether tai chi helps knee osteoarthritis. Clinical guidelines already recommend it. The interesting question is whether you can strip away the instructor, the class, and the clinic and still get a useful result.

That is what the 2026 RETREAT randomized clinical trial in JAMA Internal Medicine tested. In 178 Australian adults with knee osteoarthritis, researchers compared a 12-week unsupervised online Yang-style tai chi program with an active control: an online osteoarthritis education program. This was not tai chi versus nothing. It was tai chi versus another credible online intervention, which raises the bar.

The main result is practical: in a digitally connected group of adults with symptomatic knee osteoarthritis, a free, self-directed online tai chi program reduced pain and improved function over 12 weeks.

The pain result was the headline outcome. On a 0 to 10 numeric rating scale, pain improved by 2.7 points in the tai chi group and 1.3 points in the education group, for a between-group mean difference of -1.4 points (95% CI, -2.1 to -0.7; P<.001). Physical function, measured with WOMAC on a 0 to 68 scale, improved by 12.0 points with tai chi and 6.9 points with education, for a between-group mean difference of -5.6 points (95% CI, -9.0 to -2.3; P<.001). Those are statistically significant differences, and the responder analysis matters even more for real people: 73% of the tai chi group versus 47% of controls achieved a clinically important improvement in pain, and 72% versus 52% did so for function.

The trial also held together better than many digital health studies do. According to the primary paper, 170 of 178 participants, or 96%, completed the 12-week primary outcome assessment. The published 2024 trial protocol had already laid out the primary outcomes, sample size calculation, and analysis plan. That does not make the study flawless, but it does cut down the usual worries about outcome switching and improvised storytelling after the data arrive.

Why the result is plausible

RETREAT did not appear out of nowhere. Tai chi already had a decent evidence base in knee osteoarthritis before this online trial. A 2024 meta-analysis of mind-body exercise trials found that tai chi, yoga, and Baduanjin improved pain, stiffness, physical function, and depression in people with knee osteoarthritis. A 2025 meta-analysis of 24 randomized trials reported pooled effect sizes across low-intensity physical activity interventions of about d=-0.65 for pain, d=-0.71 for stiffness, and d=-0.58 for function. A 2021 tai chi meta-analysis found improvements in walking function and postural control.

That broader literature matters because RETREAT did not need to prove tai chi from scratch. It needed to show that online, unsupervised delivery could preserve a useful share of the benefit. On that narrower question, the trial looks credible. Its effect sizes landed in the same neighborhood as prior supervised studies, at least for short-term self-reported pain and function.

There is also a biomechanical reason the intervention is not ridiculous. A 2021 biomechanical analysis of four common tai chi movements found that they produce joint-angle and joint-moment patterns that differ from ordinary walking, with a wide range of lower-limb motion and slower increases in joint loading. That is a plausible mechanism for symptom benefit and movement training. Still, plausibility is not proof. That study described how tai chi movements load the lower limb in an experienced practitioner. It did not show that online tai chi reduces knee joint load compared with a control program in patients with knee osteoarthritis.

The same caution applies to the broader biomechanics literature. A 2023 systematic review and meta-analysis on exercise therapy and biomechanical load found that exercise can modify markers such as the first peak knee adduction moment, with effects shaped by body weight, gait speed, and walking distance. But that was exercise therapy broadly. It was not proof that tai chi is the special ingredient, and it was not proof that an unsupervised website achieves those biomechanical changes.

What the trial improved, and what it did not

RETREAT improved more than the two primary endpoints. Secondary outcomes favored tai chi for KOOS pain, sport and recreation function, quality of life, physical and mental well-being, global improvement ratings, pain self-efficacy, and balance confidence. No serious adverse events were reported. For a first-line non-drug therapy, that safety profile matters.

Some secondary outcomes did not improve. The paper did not find significant differences for fear of movement, sleep quality, positive activated affect, or oral medication use. That pattern is useful. It suggests the program was not a universal upgrade to every part of life in 12 weeks. It helped pain and function. It did not solve everything else just because it came wrapped in slow breathing and graceful arm circles.

Why the conclusion should stay modest

The cleanest way to overstate this trial would be to say online tai chi works as well as in-person tai chi. The study did not test that. It compared online tai chi with online education, not with a live supervised class. The published evidence therefore supports a narrower claim: unsupervised online tai chi beat an active online control over 12 weeks on self-reported pain and function.

That distinction matters because supervised studies can measure things RETREAT did not. An in-person 2025 pilot randomized trial reported improvements in clinical outcomes plus objective gait biomechanics and neuromuscular measures after supervised Chen-style tai chi. A 2021 randomized trial of supervised Sun-style tai chi found broader functional fitness gains in older adults with mild degenerative knee osteoarthritis. RETREAT had no six-minute walk test, no Timed Up and Go, no dynamometry, and no gait lab. If participants moved better in a measurable mechanical sense, the trial did not show it.

There is also a specific complication from the research group's own earlier work. The RETREAT protocol paper notes a predecessor online yoga trial with a similar design and 12-week duration in 212 participants. That program improved function but not pain. The contrast does not disprove RETREAT. It does mean the field still has to ask whether tai chi has some modality-specific advantage, or whether this is ordinary trial-to-trial variation in closely related digital exercise studies.

The access problem did not disappear. It moved online.

Digital delivery looks scalable on paper. Real life is less tidy. RETREAT recruited online and required internet access, a computer or tablet, and a smartphone. The participants were community-dwelling adults in Australia, with a mean age around 61 to 62 years, and women made up roughly 66% to 74% of the sample. That is not the whole knee osteoarthritis population. It skews younger than many clinical populations with severe disability, and it favors people already comfortable enough with technology to enroll in an internet-based trial.

A 2025 systematic review of digital technology in home-based exercise makes the problem plain: uptake and adherence depend heavily on digital literacy, socioeconomic status, device access, and connectivity. The people most likely to be excluded by an online exercise program are often the people with the biggest barriers to in-person care as well. Digital health can widen access, but only for those who can get through the digital front door.

Simpler options may do just as well

Tai chi has cultural prestige, a long clinical literature, and better branding than most home exercise prescriptions. None of that proves it is uniquely necessary. The 2025 meta-analysis by Qi et al. found that walking, Baduanjin, tai chi, and general physical activity all improved pain, stiffness, and function across trials. If lower-complexity options produce similar effect sizes, then the extra burden of learning a 24-form Yang-style sequence may not buy extra benefit, especially without instructor feedback.

That is not an argument against tai chi. It is an argument for comparative trials that ask the question directly. If a simpler walking program or Baduanjin routine performs just as well online, many patients and clinics would reasonably choose the easier tool.

What readers should take from this

The RETREAT trial adds something genuinely useful to the knee osteoarthritis literature. It shows that a free, self-directed online tai chi program can improve pain and function over 12 weeks compared with an active educational control, with no serious adverse events reported. The study was randomized, pre-registered, and had unusually high follow-up for a digital trial. Those are real strengths.

It also leaves large questions open. The trial did not prove equivalence to in-person tai chi. It did not measure objective functional or biomechanical outcomes. It did not report long-term follow-up beyond 12 weeks. And it tested a digitally connected sample rather than the full range of patients who struggle with knee osteoarthritis.

So the sober verdict is this: for middle-aged and older adults who can use the technology, online tai chi looks like a credible short-term symptom-management option, not a revolution and not a substitute for every other form of exercise care. Sometimes a useful result is just a useful result. Medicine could use more of those.