Exercise therapy for knee osteoarthritis: what the evidence shows

Why this question is worth asking carefully

If your knees ache going down stairs, stiffen after you've been sitting, or feel unsteady by the end of the day, you're in very large company. Knee osteoarthritis is one of the most common chronic joint conditions in the world, and "move more" is the advice almost everyone with it eventually hears. It sounds simple — but it raises a fair question: does exercise actually help the knee, or is it just generic wellness advice? Researchers have studied this closely, and the honest answer is more nuanced than either "it's a cure" or "it's pointless." This piece lays out what the best current evidence found, how confident researchers are in it, and why the numbers and the official guidelines don't line up as neatly as you might expect. We're not here to sell you on exercise or to talk you out of it — just to show you the evidence as it stands so you can have a better conversation with a regulated health professional.

What knee osteoarthritis actually is

Osteoarthritis is a degenerative joint disease in which the tissues of the joint break down over time [1]. It isn't only "worn cartilage": the damage can involve the cartilage that caps the ends of the bones, plus the tendons, ligaments, synovial lining, the bone itself, and — in the knee — the meniscus [1]. As those tissues change, people commonly notice pain, stiffness that eases within about half an hour of getting moving, swelling after activity, and a sense of the joint being less stable; some hear a grinding or scraping sound as the knee moves [1].

It is also strikingly common. The World Health Organization estimates that about 528 million people worldwide were living with osteoarthritis in 2019, and that the knee is the single most frequently affected joint, at roughly 365 million people [2]. Osteoarthritis is a major contributor to years lived with disability [2]. There is no treatment that reverses the underlying joint changes, so care focuses on managing symptoms and maintaining function — which is exactly why non-drug options like exercise draw so much attention [1][2].

What the evidence shows

The most current high-quality synthesis is a 2024 Cochrane systematic review by Lawford and colleagues, which pooled 139 randomized trials involving 12,468 people with knee osteoarthritis [3]. Cochrane reviews sit at the top tier of evidence because of how rigorously they assess each study's quality and grade their overall confidence. As with most fair tests, the picture depends heavily on what exercise is being compared against.

Compared with no treatment or usual care. This tells us how someone who takes up exercise tends to fare versus someone who carries on as usual. Here exercise looked best: immediately after the program, pain was about 13 points lower on a 0–100 scale (mean difference 13.14, 95% CI 10.36 to 15.91) and self-reported physical function was about 12.5 points better (mean difference 12.53, 95% CI 9.74 to 15.31) [3]. Cochrane rated the pain result low-certainty and the function result moderate-certainty [3].

Compared with an attention control or placebo. This is the tougher test, designed to separate the exercise itself from the attention, encouragement, and expectation that come with any hands-on program. Against that comparison the measured benefit was smaller: pain about 8.7 points lower (mean difference 8.7, 95% CI 5.7 to 11.7, low-certainty) and physical function about 11.3 points better (mean difference 11.27, 95% CI 7.64 to 15.09, moderate-certainty) [3]. Quality-of-life gains were smaller still and less certain across comparisons [3].

Here is the part a careful reader shouldn't skip. Reviewers compare results against a "minimal important difference" — the smallest change a person is likely to actually notice. The 2024 review concluded that, judged against those thresholds, the benefits were of uncertain clinical importance: several confidence intervals either fell short of the "you'd notice it" mark or straddled it, spanning both meaningful and not-meaningful improvement [3]. The authors' own summary is worth quoting: they found "low- to moderate-certainty evidence that exercise probably results in an improvement in pain, physical function, and quality of life in the short-term," but that "these benefits were of uncertain clinical importance" [3]. In plain terms: exercise probably helps in the short run, the effect is real but modest, and researchers can't be confident it crosses the line into a difference the average person clearly feels.

How long it lasts. The 2024 review focused mainly on outcomes measured right after programs ended, and longer-term data were more limited [3]. The earlier 2015 version of the same Cochrane review, by Fransen and colleagues, did report follow-up: two to six months after treatment stopped, the sustained benefit was smaller — roughly 6 points for pain (95% CI 3 to 9) and 3 points for function (95% CI 1 to 5) on the 0–100 scale [4]. The signal points one way fairly consistently: benefits tend to fade as people stop exercising, which makes durability largely a question of whether the activity continues.

Safety. Exercise was not free of downsides in the trials, though reported problems were generally minor. The 2024 review found that, depending on the comparison, exercise may slightly increase the number of reported adverse events (such as increased joint or muscle soreness), with certainty ranging from low to moderate; serious harms were not a prominent feature [3].

How it's studied — and why that matters

Three features of this research should temper how firmly anyone reads the numbers.

• "Exercise" is not one thing. The trials pooled walking programs, strengthening, aquatic classes, tai chi, cycling, and supervised physiotherapy, at different intensities, durations, and dose. Bundling them produces a useful average but hides real variation, and the review found sparse evidence that any single type clearly outperforms another [3]. An average across very different programs is a blunt instrument.
• You can't blind a person to exercise. In a drug trial you can hand out an identical dummy pill; you cannot make someone unaware they've been assigned to a walking program. Because participants know what they're getting, expectation and the attention of a supervising clinician can inflate the reported benefit — one reason Cochrane graded much of the evidence below "high" certainty [3]. The tougher attention-control comparison, which tries to hold that attention constant, is exactly where the measured effect shrank [3].
• Adherence and study quality vary. Benefits depend on people actually doing the exercise, and many trials were small or had methodological weaknesses that leave results vulnerable to bias — a further reason the certainty ratings landed at low to moderate rather than high [3].

What the guidelines say

Here's a tension worth naming honestly. Even though the current Cochrane review calls the size of the benefit "uncertain" in clinical terms, major clinical guidelines still recommend exercise strongly and consistently:

• The Osteoarthritis Research Society International (OARSI) 2019 guidelines designate structured land-based exercise as a Core Treatment for knee osteoarthritis — a foundational recommendation for essentially everyone, alongside education and weight management [5].
• The American College of Rheumatology / Arthritis Foundation 2019 guideline strongly recommends exercise for knee osteoarthritis, noting a strong body of literature supporting its use for almost all patients with the condition [6].

Why would guidelines land firmly on exercise when the effect size is modest and uncertain? Because guideline panels weigh more than a single placebo-controlled number. They factor in that osteoarthritis has no cure and no low-risk option that reverses it, that exercise is comparatively safe and inexpensive, that it carries broad benefits for heart, weight, and general health, and that it is a reasonable first-line step before or alongside medication or surgery [1][2][5][6]. A recommendation "for almost all patients" is a judgment about the overall balance of benefits, harms, and alternatives — not a claim that exercise produces a large, guaranteed change in any one person's knee. Reading it as the latter would overstate what the trials actually measured.

What this means for you

If you're living with knee osteoarthritis, a fair summary is this: the best current evidence suggests exercise probably improves pain and function in the short term, the effect is genuine but modest and of uncertain clinical importance, benefits appear to fade if you stop, the risks are generally minor, and clinical guidelines still recommend it as a core, first-line option [3][4][5][6]. None of that is a promise of relief, and none of it is a substitute for a proper assessment of your knee.

The right next step is a conversation with a licensed, regulated health professional — your physician, a physiotherapist, or another qualified clinician — who can examine your specific situation, rule out problems that need different care, and help you decide what kind of exercise, at what dose, makes sense for you, and how it fits alongside your other options. That individualized judgment is something the research averages alone cannot provide.