Condition: Arthritis & Joint Pain

Against placebo on a 10 cm visual analogue scale, the difference in pain was 0.4 cm (95% credible interval 0.7 to 0.1) for glucosamine, 0.3 cm (0.7 to 0.0) for chondroitin and 0.5 cm (0.9 to 0.0) for the combination, against a pre-specified minimal clinically important difference of 0.9 cm (0.4 in) that none of the intervals reached. Joint space narrowing did not differ. Industry-independent trials showed smaller effects than commercially funded trials, P = 0.02 for the interaction. Measured in: 10 randomized trials averaging at least 100 patients per treatment arm, 3,803 people, plus a separate review of 20 trials and 3,846 people. The two meta-analyzes are not independent corroboration: they share three authors, the same Bern methods group, and the same large-trials-only strategy, and the earlier review's trials feed the later one. The one independent leg is the 1,583-person trial. In the second review, a pooled effect of -0.75 falls to -0.03 (95% CI -0.13 to 0.07) when restricted to the three large intention-to-treat trials, which is 0.6 mm on a 10 cm scale. The network analysis notes that all but three of its included trials were manufacturer-funded, and it was itself funded by a national science foundation with no author affiliated to any manufacturer.

Pooling nine trials, arthroscopy beat control by an effect size of 0.14 (95% CI 0.03 to 0.26), which is 2.4 mm (0.4 to 4.3) on a 100 mm pain scale, present at three and six months and absent by one to two years, with no benefit on physical function (0.09, -0.05 to 0.24). Against sham surgery specifically: no advantage at any point over 24 months in 180 patients with knee osteoarthritis, and no difference on any primary outcome at 12 months in 146 patients with a degenerative meniscal tear and no osteoarthritis. Measured in: Nine randomized trials of middle-aged and older patients with knee pain and degenerative knee disease, with or without radiographic osteoarthritis; the two sham-controlled trials contributed 180 and 146 patients. None of this applies to a locking knee, a bucket-handle tear, or a traumatic tear in a younger athlete, none of which these trials recruited. Harms are not zero: symptomatic deep vein thrombosis occurred at 4.13 per 1,000 procedures (95% CI 1.78 to 9.60), alongside pulmonary embolism, infection and death.

Restricted to 24 large placebo-controlled trials in 8,997 participants, pain fell by a standardized mean difference of 0.08 (95% CI 0.15 to 0.02), about 2.0 mm on a 100 mm scale, against a prespecified minimal clinically important difference of 0.37. Trial sequential analysis dates conclusive evidence of clinical equivalence with placebo to 2009. Serious adverse events were higher than placebo, relative risk 1.49 (1.12 to 1.98), across 15 large trials in 6,462 participants. Measured in: 169 trials and 21,163 randomized participants with knee osteoarthritis; the main analysis used only trials with at least 100 participants per group. Egger's tests and asymmetric funnel plots showed clear small-study effects and publication bias across the whole literature, which is why the main analysis excludes small trials; that exclusion is the analytic choice driving the result, and it is defensible, not neutral. The serious adverse event finding is a pooled relative risk across heterogeneous products and does not identify which preparation carries it.

2024 Cochrane update, 139 trials: exercise beat an attention control or placebo by 8.70 points on pain (95% CI 5.70 to 11.70) and 11.27 on physical function (7.64 to 15.09) on 0 to 100 scales, and beat no treatment, usual care or limited education by 13.14 (10.36 to 15.91) and 12.53 (9.74 to 15.31). The 2015 version reported a standardized mean difference of 0.49 for pain, about 12 points from a control baseline of 44. No difference was found between types of exercise, and none between the number of sessions prescribed and the effect. Measured in: 139 randomized trials, 12,468 people with knee osteoarthritis (2024 update); 54 trials, about 4,600 people (2015 version). The reviewers used minimal important differences of 12 points for pain and 13 for function, and the confidence intervals either fall short of those thresholds or straddle them, so the benefit is real and of uncertain clinical importance by the review's own standard. Participants in 94% of trials were unblinded and knew they were exercising, which is part of what the self-reported pain scores measure. Adverse events against no treatment ran at a risk ratio of 3.17 (1.17 to 8.57), low certainty, and were overwhelmingly increased knee or back pain.

Pain standardized mean difference 0.38 (95% CI 0.20 to 0.55) and physical function 0.38 (0.05 to 0.54) immediately after treatment, equivalent to 8 points on pain and 7 on function from a control baseline of 29 on a 0 to 100 scale, number needed to treat 6 for each. The reduction held at three to six months after supervised treatment ended. Measured in: 9 to 10 randomized trials, 549 people with symptomatic hip osteoarthritis. This evidence base is a twentieth the size of the knee one, and only five of the ten trials recruited hip osteoarthritis exclusively. No trial blinded participants, and pain, function and quality of life were all self-reported. Quality of life was evaluated in three small studies with no measurable improvement.

In 12 knee osteoarthritis trials providing two to six month post-treatment data on 1,468 people, the pain benefit fell from a standardized mean difference of 0.49 immediately after treatment to 0.24 afterwards, about 6 points on a 0 to 100 scale. Physical function fell from 0.52 to 0.15, about 3 points. Measured in: 12 trials, 1,468 participants for pain and 1,279 for function, all with knee osteoarthritis. These follow-up subsets are not the whole review, and trials that keep participants long enough to measure sustainability differ from those that do not. The reviews record what was prescribed, not what participants continued doing, so this halving cannot be attributed to adherence specifically even though adherence is the obvious candidate. The knee-pain estimate rests on 12 trials and 1,468 people; the physical-function estimate on 10 trials and 1,279.

Sorting IDEA trial participants by weight lost over 18 months gave a significant dose-response across pain (P = 0.01), function (P = 0.0006), six-minute walk distance (P < 0.0001), physical and mental quality of life, knee joint compressive force and interleukin-6. Losing 20% or more of body weight produced 25% less pain and better function than losing 10 to 20%. In the main trial, diet plus exercise produced 23 lb (10.6 kg) of loss and less pain (3.6 on a 0 to 20 scale) than diet alone (4.8) or exercise alone (4.7). Measured in: 240 overweight and obese community-dwelling adults aged 55 and over with painful radiographic knee osteoarthritis, from the 454-participant IDEA randomized trial. The dose-response analysis is a secondary analysis in which participants were grouped after the fact by an outcome they achieved, so the group losing 20% or more may differ from the others in motivation, baseline health and comorbidity as well as in weight. A 20% loss sustained over 18 months was delivered with dietitian contact, partial meal replacement and supervised exercise, which is a larger intervention than the advice most people receive.

After 12 months and three injections, knee pain changed by 2.1 points with PRP and 1.8 with saline placebo (difference 0.4, 95% CI -0.9 to 0.2, P = 0.17). Medial tibial cartilage volume changed by 1.4% and 1.2% (difference 0.2%, -1.9% to 1.5%). Of 31 prespecified secondary outcomes, 29 showed no between-group difference. Measured in: 288 adults with symptomatic mild to moderate radiographic knee osteoarthritis, mean age 61.9, 169 (59%) women, 93% completing. One PRP preparation and one injection schedule were tested, and PRP products differ substantially in platelet concentration, leukocyte content and activation method, so this trial constrains, not closes the question. Two of the trial's authors provide PRP injections in clinical practice, declared in the paper. Participants had mild to moderate disease; severe radiographic osteoarthritis was not studied.

Over two years, serum 25-hydroxyvitamin D rose by 16.1 ng/mL on treatment against 2.1 on placebo. Knee pain fell by 2.31 on treatment and 1.46 on placebo, with no significant difference at any time point. Cartilage volume fell by 4.30% and 4.25% respectively (P = 0.96). No secondary clinical endpoint differed. Measured in: 146 adults with symptomatic knee osteoarthritis, 85% completing two years. Baseline knee pain and function were worse in the treatment group, which complicates the comparison in the direction of making treatment look worse. The trial tested supplementation to a target level, not correction of deficiency in a deficient population, so it does not address whether someone who is truly deficient would respond.

Over six to 12 weeks, about 60% of participants had much reduced pain. Number needed to treat for clinical success was 9.8 (95% CI 7.1 to 16) for topical diclofenac across six trials in 2,343 participants and 6.9 (5.4 to 9.3) for topical ketoprofen across four trials in 2,573. Where topical and oral NSAIDs were compared directly, efficacy was similar. Systemic adverse events did not differ from the carrier gel; topical diclofenac raised mild local skin reactions. Measured in: 39 studies, 10,857 participants, almost all with knee osteoarthritis. The efficacy results come almost entirely from knee osteoarthritis, so they do not transfer to a hip, which sits too deep for topical delivery. Clinical success on the carrier gel alone occurred in around half of participants over six to 12 weeks, roughly twice the rate seen with oral placebo, so part of what a topical preparation delivers is the rubbing and the vehicle. Up to 6,000 participants' worth of completed unpublished data was unavailable to the reviewers.

Statistically significant, moderate benefits on pain, function and quality of life for up to 13 weeks. Gastrointestinal adverse events were three to four times more common than placebo. Measured in: Seven randomized trials (2,102 participants) reviewed, five (1,713 participants) pooled, in knee osteoarthritis. Nothing here runs beyond 13 weeks, and osteoarthritis is a decades-long condition, so the durability is unmeasured. Duloxetine has a discontinuation syndrome that requires tapering, and roughly a fifth of adults with osteoarthritis carry a concurrent depression diagnosis, which makes the pain-specific component of the benefit hard to isolate in this population.

Pain benefit against control was a standardized mean difference of 0.48 (95% CI 0.27 to 0.70) at one to two weeks, 0.41 (0.21 to 0.61) at four to six weeks, 0.22 (0.00 to 0.44) at 13 weeks, and 0.07 (-0.11 to 0.25) at 26 weeks. The overall pain estimate of 0.40 corresponds to 1.0 cm on a 10 cm scale, number needed to treat 8. Function improved by 0.33, number needed to treat 10. Measured in: 27 randomized trials, 1,767 participants with knee osteoarthritis. The reviewers graded all outcomes low certainty: heterogeneity ran at 68%, the funnel plot was asymmetric, and most trials carried high or unclear risk of bias. Treatment effects were smaller in the trials that randomized at least 50 or 100 participants per group, which is the standard signature of small-study bias.

Against a waiting-list control, acupuncture gave a standardized mean difference of 0.96 (95% CI 0.72 to 1.19) for pain, 14.5 points on a 100 point scale, and 0.89 (0.60 to 1.18) for function, 13.0 points, across four trials in 884 participants. The reviewers judged both clinically relevant. Measured in: Four waiting-list-controlled trials, 884 participants, within a review of 16 trials and 3,498 people with knee or hip osteoarthritis. A waiting-list control receives nothing at all: no practitioner, no appointment, no attention, no expectation of improvement. The gap between this figure and the sham-controlled figure in the same review is the measure of how much of that is doing the work, and it is most of it.

Cochrane found a standardized mean difference of 0.28 (95% CI 0.11 to 0.45) for pain across nine sham-controlled trials in 1,835 people, 0.9 points on a 20 point scale, below the reviewers' 1.3 point threshold for clinical relevance; restricting to the shams most likely to blind participants made the pooled benefit smaller and non-significant, and at six months it was 0.10 (-0.01 to 0.21). A 282-person trial found neither needle nor laser acupuncture better than sham at 12 weeks or one year. A 480-person Chinese trial of intensive electroacupuncture found response rates of 60.3% against 47.3% for sham, a 13.0 percentage point gap that held to week 26. The individual patient data meta-analysis across chronic pain puts the sham margin at about 0.2 standard deviations. Measured in: Nine sham-controlled osteoarthritis trials in 1,835 people; a 282-person Australian trial of adults over 50; a 480-person Chinese trial; and 39 trials in 20,827 patients across four chronic pain conditions. Direction is marked as mixed because these are not reconcilable into one number: the pooled estimate is positive and below the threshold its own reviewers set, the largest Western trial found nothing against sham, and the largest high-dose Chinese trial found a clear gap. Sham acupuncture is itself an intervention involving a practitioner, a room and half an hour of attention, and trials using penetrating needles for sham produce smaller differences than those using non-penetrating ones.

Combining the two sham-controlled trials, pain differed by a standardized mean difference of 0.13 (95% CI -0.49 to 0.22), 2.1 points on a 100 point scale, graded moderate-quality evidence of little or no effect. Function was 0.15 (-0.51 to 0.21). Measured in: Two sham-controlled trials, 120 participants, within a review of six trials and 413 people with hip osteoarthritis, mean age 61 to 67, about two thirds women. 120 people is small enough that the confidence interval includes both a moderate benefit and nothing at all, so this is an absence of demonstrated effect at this sample size, not a demonstrated absence. Both shams were judged at risk of carrying weak acupuncture-specific effects: one placed non-penetrating needles at correct points, the other penetrating needles at incorrect ones. All results were short term, four to nine weeks.

Pooled across trials, turmeric or curcumin at about 1,000 mg/day reduced pain on a visual analogue scale by 2.04 points (95% CI 1.24 to 2.85) against placebo and WOMAC by 15.36 (3.77 to 26.9), with no significant difference against pain medication. In a separate 367-person trial it met non-inferiority against ibuprofen on WOMAC total, pain and function, though not on stiffness, which reached only P=0.060, with less abdominal discomfort. Measured in: Eight randomized trials in the meta-analysis, with individual trials ranging from about 40 to 367 participants with knee osteoarthritis or arthritis. The reviewers state their trial count, sample size and methodological quality were insufficient for definitive conclusions. The first author of that review is the president of a dietary supplement manufacturer. The comparator trial was part-funded by a state pharmaceutical manufacturer that produces turmeric capsules, and it is one of the eight trials inside the meta-analysis quoted alongside it, not an independent confirmation of it. The absorption-enhancement strategies that make these preparations work are also the ones appearing in the liver injury reports below.

Among confirmed cases, joint pain had a sensitivity of 85% (95% CI 78 to 90), a history of joint swelling 78% (71 to 85) and fever 57% (52 to 62). Sweats occurred in 27% and rigors in 19%. The strongest available discriminator is the synovial fluid white cell count: likelihood ratio 0.32 below 25,000/microliter, 2.9 at or above 25,000, 7.7 above 50,000 and 28.0 above 100,000. Age, diabetes, rheumatoid arthritis, joint surgery, a hip or knee prosthesis, skin infection and HIV all raise the probability. Measured in: 14 studies, 6,242 patients presenting with a painful swollen joint, of whom 653 met the gold standard for septic arthritis. These figures come from patients who reached the point of joint aspiration, which is a selected group already suspected of infection, so the sensitivities do not transfer directly to an unselected primary care population. The review predates routine synovial fluid PCR and covers non-gonococcal bacterial arthritis specifically.

A reduction of 2.2 lb (1 kg) in body mass was associated with a 40.6 N reduction in peak knee compressive force and a 38.7 N reduction in resultant force, about a fourfold reduction in joint load for each unit of weight lost, plus a 1.4% reduction in knee abduction moment. Measured in: 142 sedentary, overweight and obese older adults with self-reported disability and radiographic knee osteoarthritis, studied by three-dimensional gait analysis. What could explain it instead: People who lose weight also change how much and how fast they walk, and gait speed is itself a determinant of peak knee force. Some of the association between lower follow-up body mass and lower joint load may reflect the altered walking pattern of someone who has been dieting and exercising, not the mass alone.. This is a regression on gait laboratory measurements, not a trial, and it measures force during controlled walking, not during the varied loading of a real day. It establishes a mechanism by which weight loss could reduce pain and does not by itself show that pain falls.

11.5% (85/738) of the Leiden cohort and 5.4% (29/533) of the French ESPOIR cohort achieved DMARD-free sustained remission. The relationship between symptom duration and remission was not linear in either cohort; the symptom duration with optimal discriminative ability was 14.9 weeks (95% CI 12.3 to 16.0) in Leiden and 19.1 weeks (12.3 to 28.0) in ESPOIR, and 11.4 weeks for ACPA-positive disease. Measured in: 738 patients in the Leiden Early Arthritis Clinic and 533 in the French ESPOIR cohort, all with rheumatoid arthritis. What could explain it instead: Confounding by disease severity, running in both directions. People with abrupt, severe, highly inflammatory onset present sooner and also have more aggressive disease; people with insidious onset present later and may have milder disease that would remit anyway. Neither cohort randomized the delay, so part of the association between short symptom duration and later remission reflects who arrives quickly, not what earlier treatment does.. Area under the curve was 0.56 to 0.61, meaning symptom duration discriminates only weakly at the individual level; this supports the existence of a confined window, not a precise deadline. Drug-free sustained remission is a strict outcome achieved by a minority, and the study cannot say whether the same window applies to less demanding outcomes such as low disease activity on treatment.

Median lag from disease onset to the first rheumatology assessment was 1.00 years (IQR 0.5 to 2). 30% were seen within six months, 53% within one year and 71% within two years. On stepwise regression, seeing a rheumatologist later than six months was associated with peripheral joint erosions (odds ratio 4.25, P = 0.001) and worse Health Assessment Questionnaire scores (odds ratio 2.2, P = 0.004). Low education status (OR 2.09) and lower body mass index (OR 0.92) predicted a delay beyond two years. Measured in: 283 patients with psoriatic arthritis in a single cohort. What could explain it instead: Disease phenotype at onset. Psoriatic arthritis that begins as an obviously swollen, painful, rapidly progressive joint gets referred quickly; disease that begins as vague aching gets referred late. If the two phenotypes also differ in their long-run erosive potential, then part of the erosion difference is the disease sorting itself, not the wait.. Observational and single-cohort, so it establishes that late presenters do worse and not that the delay itself caused it. The regression adjusts for a limited set of variables, and a six-month threshold chosen within the same dataset that tests it is prone to overfitting.

At 24 weeks, 65.0% (95% CI 51.6 to 76.9) of the Tripterygium group and 32.8% (21.3 to 46.0) of the sulfasalazine group met ACR20, P = 0.001, in a mixed model imputing for dropouts. ACR50 and ACR70 also favored Tripterygium. Interleukin-6 fell rapidly in the Tripterygium group. Radiographic progression was lower but not significantly. Adverse event frequency was similar between groups. Measured in: 121 patients with active rheumatoid arthritis and six or more painful and swollen joints, across two US academic centers and nine rheumatology clinics, taking 60 mg of extract three times daily. Outcome data were available for only 62 of 121 patients at 24 weeks, and 38% of the Tripterygium arm and 59% of the sulfasalazine arm discontinued, so the reported figures rest on imputation over more missing data than present data. Sulfasalazine is a weak comparator by current standards; methotrexate is the reference treatment. The trial did not measure the harms this plant is best known for: amenorrhea and reduced fertility, marrow suppression and hepatotoxicity, which are the reasons it is not a self-treatment.