Osteoarthritis is the most common joint disease. It affects an estimated more than 240 million people worldwide, including more than 32 million in the United States, and it is the most frequent reason for activity limitation in adults [1]. The Global Burden of Disease programme has mapped its prevalence across 204 countries and projected it forward to 2050 [2].
It is usually described as "wear and tear", and that framing is wrong in a way that matters. Osteoarthritis involves pathological change in cartilage, bone, synovium, ligament, muscle and periarticular fat — the whole joint as an organ, not a bearing surface wearing out [1]. Cartilage is lost and its biochemistry is abnormal [3], bone remodels and forms osteophytes, and there is genuine low-grade inflammation: synovitis is prevalent at all stages, relates to pain and poor function, and may independently drive both the onset and the structural progression of radiographic disease [4] [5] [6]. The point has been made with some force in the literature — osteoarthritis is not osteoarthrosis [7] — and the inflammatory and anti-inflammatory cytokine biology is well characterised [8] [9] [10].
The strongest argument against "wear and tear" is archaeological. Comparing skeletal samples, knee osteoarthritis prevalence was 16 percent in a post-industrial sample against 6 percent in an early-industrial sample and 8 percent in a prehistoric one — and after controlling for age, BMI and other variables it remained 2.1-fold higher (95% CI 1.5–3.1) in the post-industrial group. Longevity and body weight are insufficient to explain the roughly doubled prevalence since the mid-twentieth century, which means knee osteoarthritis is more preventable than is commonly assumed [11]. Something about modern life is doing this, beyond simply living longer and heavier.
How it differs from the other two arthritides. Rheumatoid arthritis is autoimmune: the immune system attacks the joint lining, and the treatment suppresses immunity. Gout is a crystal disease: urate exceeds its solubility, and the treatment dissolves the crystals. Osteoarthritis has neither an autoantigen nor a crystal. It has no disease-modifying drug [12], and that is the central honest fact of this review.
Start here: what actually drives it
Obesity is the largest modifiable risk factor, and it has been quantified at scale: in a population-based cohort using primary-care records covering more than 5.5 million people in Catalonia, 1,764,061 subjects aged 40 and over without osteoarthritis were followed for incident clinically diagnosed knee, hip and hand disease by WHO BMI category [13].
The mechanism is not only mechanical. Dynamic load at baseline predicts radiographic disease progression in medial compartment knee osteoarthritis [14], so mechanics matter — but hand osteoarthritis is also associated with obesity, and hands do not bear weight. Adipose tissue is metabolically active and inflammatory, and the link between obesity and osteoarthritis is more than just wear and tear [15] [16]. Occupational ergonomic exposure is a further independent contributor [17], and major joint injury is among the established risk factors alongside age and female sex [1].
The consequences run beyond the joint. People with osteoarthritis have more comorbidities and are more sedentary than those without, and the reduced physical activity is associated with a 20 percent higher age-adjusted mortality [1].
Pillar 1: measurement and diagnosis
The diagnosis is clinical
Osteoarthritis is diagnosed from the pattern: joint pain, brief morning stiffness, crepitus and bony enlargement. The brevity of the stiffness is the useful discriminator, and it is operationalised in trial entry criteria — a knee osteoarthritis study required morning stiffness lasting less than 30 minutes together with knee crepitus and radiographic confirmation [18], where a rheumatoid arthritis cohort used morning stiffness of 45 minutes or more among its criteria [19]. Several physical examination findings are diagnostically useful, including bony enlargement in knee osteoarthritis and pain on internal rotation of the hip [1], and work continues on classification criteria for early disease [20]. Symptoms are usually measured with the WOMAC index, a validated instrument for pain, stiffness and function [21].
The X-ray, and what it grades
Radiographs show joint-space narrowing, osteophytes and subchondral sclerosis, graded 0 to 4 on the Kellgren-Lawrence scale [22], with an atlas of individual radiographic features for standardisation [23]. MRI can score the whole organ [24], and there are standard recommendations for histological assessment [25].
And then there is the problem that organises the rest of this review.
Centerpiece: two measurements that barely overlap
The Johnston County Osteoarthritis Project examined 3,068 adults aged 45 and over and reported four prevalences side by side: hip symptoms in 36 percent; radiographic hip osteoarthritis (Kellgren-Lawrence grade 2 or above) in 28 percent; symptomatic hip osteoarthritis — symptoms in a hip that also has radiographic disease — in nearly 10 percent; and moderate-to-severe radiographic disease (grades 3 and 4) in 2.5 percent [26].
The check the arithmetic was never given. Those three numbers sit in one abstract and are never divided by each other. Doing the division gives the discordance in both directions at once. Of the 28 percent with radiographic hip osteoarthritis, only 10 percent are symptomatic — so just 36 percent of radiographic hip osteoarthritis hurts, and 64 percent is silent. And of the 36 percent with hip symptoms, only 10 percent have radiographic disease — so only 28 percent of hip symptoms are explained by the X-ray, and 72 percent are not. The population splits into 26 percent with pain and a clean film, 10 percent with both, and 18 percent with an arthritic film and no pain.
A second check, across independent cohorts. The Framingham Osteoarthritis Study radiographed 1,424 people aged 63 to 94 in 1987 and reported radiographic knee osteoarthritis in 34 percent of women and 31 percent of men, with symptomatic disease in 11 percent of women and 7 percent of men [27]. Those give symptomatic fractions of 32.4 percent and 22.6 percent — against Johnston County's 35.7 percent for hips. Two cohorts, two different joints, twenty-two years apart, neither comparing itself to the other and neither computing this ratio, and all three land between a quarter and a third.
One more number worth stating. Moderate-to-severe radiographic disease is 2.5 percent of the population, against 10 percent with symptomatic osteoarthritis — so severe-looking films account for only about a quarter of the symptomatic group [26]. Most people with osteoarthritis pain do not have a dramatic X-ray.
The teaching point. Radiographs and symptoms are measuring different things, so osteoarthritis is managed by symptoms and function, not by imaging severity. A film cannot tell you who is suffering, and it should not be the thing that drives escalation. This is also why the treatments that work best are the ones aimed at function — and why the fact that 33 percent of people over 75 have both symptomatic and radiographic knee osteoarthritis [1] is the relevant number for planning services, rather than the radiographic prevalence alone.
Three honest limits. The Johnston County overlap is reported as "nearly 10 percent", so the derived percentages are approximate at the first decimal. Both cohorts are cross-sectional, so this is discordance at a moment, not a claim about whether radiographic change precedes or follows pain. And the two studies define symptoms slightly differently, which is part of why the three figures are not identical — the point is the band they fall in, not their exact agreement.
One further gap: this substrate, assembled by citation crawl during an OpenAlex outage, contains no study explicitly designed to measure radiographic-symptom discordance. The figure is built by dividing prevalences the source papers report but do not divide, which is legitimate arithmetic on their published numbers and is not a substitute for a study designed for the question.
Pillar 2: treatment
There is no disease-modifying drug. Management is symptom and function focused, and the guidelines converge closely: OARSI for non-surgical knee management [28], the American College of Rheumatology in 2012 and again in 2019 [29] [30] [31], and EULAR for knee and hand disease [32] [33].
The core is not a drug
Exercise and weight loss are the strongly recommended foundation. The 2019 ACR/Arthritis Foundation guideline makes strong recommendations for exercise, for weight loss in overweight or obese patients with knee and/or hip osteoarthritis, for self-efficacy and self-management programmes, for tai chi, for cane use, for hand orthoses for the first carpometacarpal joint, and for tibiofemoral bracing [30] [31]. The 2012 recommendations similarly put aerobic, aquatic and resistance exercise plus weight loss at the centre [29], as does the 2021 clinical review [1]. Patient beliefs about exercise are themselves a target, since fear of causing damage is a barrier to the treatment that works [34] [35].
And the weight-loss lever now has a drug behind it. A 68-week double-blind randomised trial at 61 sites in 11 countries assigned participants with obesity (BMI ≥ 30) and moderate knee osteoarthritis with at least moderate pain to once-weekly semaglutide 2.4 mg or placebo, on top of physical-activity counselling and a reduced-calorie diet, with co-primary endpoints of percentage weight change and change in WOMAC pain [36]. That trial matters less as a new osteoarthritis drug than as a demonstration that the most effective modifiable lever is now pharmacologically reachable.
Analgesia
Topical and oral NSAIDs complement the core in patients without contraindications [1]. The gastrointestinal cost is real and was quantified in two large trials comparing COX-2-selective agents against conventional NSAIDs — rofecoxib against naproxen [37] and celecoxib against non-selective NSAIDs in osteoarthritis and rheumatoid arthritis [38] — and the selective agents have been reviewed as a class [39].
Opioids should not be used for this. A pragmatic 12-month randomised trial in patients with moderate-to-severe chronic back pain or hip or knee osteoarthritis pain despite analgesics compared an opioid-based treat-to-target strategy against a non-opioid one, in 240 randomised patients [40]. Both arms escalated toward the same target; the comparison is therefore about the drug class rather than about treatment intensity, and it is the trial to cite when a patient asks why they are not being offered something stronger.
Injections
Intra-articular corticosteroid provides short-term pain relief [1], assessed in a Cochrane review of 28 trials and 1,973 participants [41]. The benefit is real and it is measured in weeks.
Hyaluronic acid (viscosupplementation) is the contested one. A meta-analysis identified 89 trials involving 12,667 adults, 68 of them sham-controlled [42], and the intervention has been reviewed repeatedly [43] [44] [45]. Platelet-rich plasma has been studied on the same model [46]. The guidelines above are the place to look for the current position rather than any single trial.
Glucosamine and chondroitin have a clean answer, and it is negative. A network meta-analysis of 10 large randomised trials in 3,803 patients prespecified a minimal clinically important difference of −0.9 cm on a 10 cm pain visual analogue scale, and found the difference against placebo for glucosamine was −0.4 cm (95% credible interval −0.7 to −0.1) [47]. Statistically distinguishable from zero; less than half the size of the smallest difference the authors had defined in advance as mattering to a patient. A later multicentre trial of the combination reported separately [48].
Surgery
Arthroscopy for osteoarthritis does not work, and the trial that showed it is a landmark. 180 patients were randomised to arthroscopic débridement, arthroscopic lavage, or placebo surgery — skin incisions and a simulated débridement without inserting the arthroscope — with patients and outcome assessors blinded, and followed for 24 months on five self-reported pain and function scales plus an objective walking and stair-climbing test. At no point did either intervention group report less pain or better function than the placebo group [49].
Joint replacement, by contrast, works. For people with advanced symptoms and structural damage, total joint replacement effectively relieves pain [1], and it is described as highly effective for late-stage hip and knee disease [2] [50] — with the caveat that rising demand may put it out of reach in some regions and widen health inequity [2]. What is less appreciated is what determines whether patients are satisfied with it. In a prospective cohort of 4,709 primary lower-limb joint replacements, overall satisfaction was predicted by meeting preoperative expectations (OR 2.62, 95% CI 2.24–3.07), satisfaction with pain relief (2.40), satisfaction with the hospital experience (1.7), and the 12-month and preoperative Oxford scores [51]. Expectation-setting outranked everything except pain relief itself, which is an argument for preoperative education [52]. The long-term failure mode is aseptic loosening and periprosthetic osteolysis [53].
Pillar 3: what is unresolved
The DMOAD graveyard. No drug has been shown to modify the disease, and the development record has been reviewed as such [12]. Sprifermin is the instructive case: the FORWARD trial randomised 549 patients with symptomatic Kellgren-Lawrence grade 2–3 knee osteoarthritis across five arms of intra-articular sprifermin or placebo over five years, with a primary endpoint of change in total femorotibial cartilage thickness [54]. Note what that endpoint is. After a review in which the central fact is that cartilage appearance and patient pain come apart, a trial powered on cartilage thickness is measuring the thing the figure above says is only loosely connected to how people feel. Mesenchymal stromal cell therapy for cartilage regeneration has been assessed similarly [55], and biomarkers have been proposed to support drug development [56].
Phenotypes. If osteoarthritis is several diseases wearing one name — inflammatory, metabolic, post-traumatic, bone-driven — then trials that mix them will dilute any real effect. Clinical phenotypes in knee osteoarthritis have been systematically reviewed [57] [58], and this is the most credible explanation on offer for why the DMOAD trials keep failing.
Mechanism. Chondrocyte biology gives real targets — hedgehog signalling regulates chondrocyte differentiation and is implicated in osteoarthritis [59], β-catenin activation in articular chondrocytes produces an osteoarthritis-like phenotype [60], and the pathogenic signalling pathways have been catalogued [10] [9]. Whether any is druggable in a way that changes symptoms remains open.
Prevention. The archaeological finding that prevalence has doubled beyond what age and weight explain [11] implies modifiable causes not yet identified. Weight and joint injury are the two known levers [13] [1]; the rest is unaccounted for.
Dig deeper in lmmol
Reading the arthritis trio together is the point. Rheumatoid arthritis is the autoimmune member — a disease with an identified immune mechanism, disease-modifying drugs and a treat-to-target strategy, all three of which osteoarthritis lacks; the NSAID toxicity trials cited here enrolled both conditions [38] [37], and the contrast between prolonged and brief morning stiffness is a bedside discriminator. Gout is the crystal member, and the sharpest contrast of all: gout is curable because its cause is a solubility threshold you can measure and cross, while osteoarthritis has no equivalent number. Obesity is the shared modifiable risk and now the shared drug target, since the semaglutide trial treats knee osteoarthritis by treating weight [36] [13]. Osteoporosis is worth reading alongside for the contrast in the other direction — another age-related bone and joint condition, but one where imaging does predict the outcome that matters. And depression belongs here because osteoarthritis is a chronic pain condition, and the treatments that work are behavioural as much as pharmacological [30]. The full collection is at health.