Bones and joints

What research found here. What was shown to help comes first, then the risks reported, most common first.

Bones and joints

Strong several reviews agree · Moderate one review or consistent trials · Limited one or a few small studies · Untested no study in people, listed with what it rests on

Shown to help106

25-hydroxyvitamin DStrong

Broken bones

Linked to more hip fractures at low levels in cohorts, while pooled trials of vitamin D supplements found no reduction in fractures or falls.

The numbers and sources

Effect: In 15 prospective cohort studies, the lowest category of 25-hydroxyvitamin D had a relative risk of hip fracture of 1.58 against the highest (95% CI 1.41 to 1.77), with risk falling as levels rose up to about 24 ng/mL (60 nmol/L) and flat above that [12]. In 81 randomized trials, vitamin D had no effect on total fracture (RR 1.00, 95% CI 0.93 to 1.07), hip fracture (RR 1.11, 95% CI 0.97 to 1.26) or falls (RR 0.97, 95% CI 0.93 to 1.02), and no clinically relevant effect on bone mineral density [13]. In a trial of 2,000 IU a day, total fractures did not differ from placebo (HR 0.98, 95% CI 0.89 to 1.08), with no difference by starting 25-hydroxyvitamin D level [14].
Population: 51,239 participants with 3,386 hip fractures in 15 cohort studies [12]. 53,537 participants in 81 trials [13]. 25,871 generally healthy US adults followed a median 5.3 years [14].
Sources: [12], [13], [14]
Caveats: The cohort finding is observational. The trial participants were not selected for vitamin D deficiency, low bone mass or osteoporosis [14], so the trials say less about people with very low levels.

AcetaminophenStrong

Knee and hip osteoarthritis pain

Shown to lower knee and hip osteoarthritis pain by about 3 points on a 100-point scale, below the 9 points the reviewers set as clinically important.

The numbers and sources

Effect: In 10 placebo-controlled trials, pain fell 23 points on a 0 to 100 scale on placebo and 3.23 points more on paracetamol (95% CI 1.02 to 5.43), against a minimal clinically important difference of 9. Physical function improved 2.9 points more than on placebo (95% CI 0.95 to 4.89). Effects did not differ between 3.0 g a day or less and 3.9 g or more [8]. An earlier meta-analysis found a similar effect on pain (-3.7 points, 95% CI -5.5 to -1.9) and disability (-2.9, 95% CI -4.9 to -0.9) [9]. An overview of systematic reviews put the pain effect at 0.3 points on a 0 to 10 scale (95% CI 0.1 to 0.6) [5].
Population: 3,541 people with hip or knee osteoarthritis in 10 trials at 1.95 to 4 g a day, most followed for 3 months [8]; 13 randomized trials in spinal pain and osteoarthritis [9].
Sources: [8], [9], [5]
Caveats: The reviews share trials and authors. No trial reported quality of life [8].

AcupunctureStrong

Chronic pain across conditions

Shown to ease chronic pain about 0.5 standard deviations more than no acupuncture and about 0.2 more than sham acupuncture, with effects lasting a year.

Compared with sham acupuncture [2][3]; no acupuncture [2][3]

The numbers and sources

Effect: In individual patient data from 39 trials, acupuncture was superior to both sham and no acupuncture for each pain condition (all P < .001). The difference was close to 0.5 standard deviations against no acupuncture and close to 0.2 against sham, and about 15% of the effect was lost at 1 year. Effects were smaller in trials whose sham used a penetrating needle [2]. The earlier analysis of 29 trials found pain scores lower than sham by 0.23 SD for back and neck pain, 0.16 SD for osteoarthritis and 0.15 SD for chronic headache, and lower than no acupuncture by 0.55, 0.57 and 0.42 SD [3].
Population: 20,827 patients with nonspecific musculoskeletal pain, osteoarthritis, chronic headache or shoulder pain in 39 trials [2]. 17,922 patients in 29 trials [3].
Sources: [2], [3]
Caveats: The two analyses come from the same collaboration and the later one contains the earlier trials. Patients and acupuncturists cannot be blinded when the comparison is no acupuncture, so the larger effect includes expectation. The authors say factors beyond needling contribute to the effect [3].

AcupunctureStrong

Knee osteoarthritis, compared with no acupuncture or a waiting list

Shown to ease osteoarthritis pain 14.5 points on a 100-point scale more than a waiting list in the short term. One later trial found no gain at 1 year.

Compared with waiting list [8]; no acupuncture [3][9]

The numbers and sources

Effect: Against a waiting list, pain improved by 14.5 points more on a 100-point scale (SMD -0.96, 95% CI -1.19 to -0.72, 4 trials, 884 participants) and function by 13.0 points (SMD -0.89) [8]. Individual patient data put the difference from no acupuncture at 0.57 SD for osteoarthritis [3]. Against no acupuncture, needle acupuncture lowered knee pain by 1.1 units on a 0 to 10 scale at 12 weeks (95% CI -1.8 to -0.4), and not at 1 year [9].
Population: People with knee or hip osteoarthritis in 4 waiting-list trials [8]. 282 adults aged 50 or older [9].
Sources: [8], [3], [9]
Caveats: None of these comparisons could blind participants, so expectation and attention are part of the effect. The 1-point change at 12 weeks was below that trial's 1.8-unit threshold for clinical importance [9].

BisphosphonatesStrongStudied in women

Fractures in postmenopausal women with osteoporosis

Shown to cut fractures in postmenopausal women with osteoporosis, with clinical spine fractures about 55% lower on alendronate, or 2.7 fewer per 100 women.

Compared with placebo or no treatment [1]; placebo [2][3][4]; placebo and other osteoporosis drugs [5]

The numbers and sources

Effect: In higher-risk women, alendronate lowered clinical vertebral fractures (RR 0.45, 95% CI 0.28 to 0.73, 2.7% fewer), non-vertebral fractures (RR 0.80, 2.8% fewer), hip fractures (RR 0.49, 1.0% fewer) and wrist fractures (RR 0.54, 1.8% fewer) [1]. Risedronate 5 mg a day lowered non-vertebral fractures (RR 0.80, 95% CI 0.72 to 0.90, 2% fewer) and hip fractures (RR 0.73, 1% fewer) [2]. Yearly zoledronic acid lowered spine fractures seen on x-ray from 10.9% to 3.3% over 3 years (RR 0.30, 95% CI 0.24 to 0.38) and hip fractures from 2.5% to 1.4% (HR 0.59), with 25% fewer non-vertebral fractures [3]. Daily and intermittent ibandronate lowered new spine fractures by 62% and 50% [4]. In a network meta-analysis, alendronate, risedronate and zoledronate each lowered hip, spine and non-vertebral fractures, while ibandronate lowered spine fractures only, with lower efficacy [5].
Population: 7,375 higher-risk women in 20 alendronate trials of 1 to 3 years [1]; 14,354 higher-risk women in 9 risedronate trials [2]; 7,765 women, mean age 73 [3]; 2,946 women [4]; 193,987 postmenopausal women in 107 trials [5].
Sources: [1], [2], [3], [4], [5]
Caveats: Many trials in [1] and [2] had unclear or high risk of bias, and most certainty ratings were low to moderate. Risedronate's effect on clinical spine fractures could not be estimated because no events were reported [2].

BisphosphonatesStrong

Bone density in people taking steroid tablets

Shown to raise spine bone density more than placebo or calcium and vitamin D in people on steroid tablets, though less than teriparatide. Trials were too small to judge fractures.

Compared with placebo or calcium and vitamin D [10]; teriparatide and denosumab [10][11]

The numbers and sources

Effect: Every osteoporosis drug raised lumbar spine bone density more than placebo or calcium and vitamin D. Teriparatide raised it more than alendronate (effect size 3.8), risedronate (4.0) and zoledronate (2.4), and zoledronate and denosumab more than oral bisphosphonates [10]. A second network meta-analysis found alendronate, risedronate, etidronate and zoledronate all raised bone density, with teriparatide and denosumab ahead at the spine and femoral neck [11].
Population: 5,260 adults on oral glucocorticoids in 31 studies, 72.5% women [10]; 2,877 patients in 11 trials [11].
Sources: [10], [11]
Caveats: Bone density is a stand-in for fractures, and the included studies were underpowered for fractures [10].

BoswelliaStrong

Knee osteoarthritis pain and function

Shown to lower knee osteoarthritis pain and stiffness and improve function against control in adults, in mostly small, low-quality trials.

The numbers and sources

Effect: In 7 trials (545 patients), boswellia lowered VAS pain (WMD -8.33, 95% CI -11.19 to -5.46) and WOMAC pain (WMD -14.22), stiffness (WMD -10.04) and function scores (WMD -10.75) against control [1]. A meta-analysis of 11 RCTs (1,009 patients) of curcuminoids and boswellia found both improved pain and function more than placebo [2]. A 2026 network meta-analysis of 20 RCTs (1,633 participants) found modified boswellia formulations improved WOMAC pain, stiffness and function [3].
Population: Adults with knee osteoarthritis, mostly mild to moderate [1][2][3].
Sources: [1], [2], [3]
Caveats: Study quality was low overall, and most trials had fewer than 100 participants. No RCT compared boswellia with approved NSAIDs [2]. The abstracts do not report trial funding.

CollagenStrong

Tendon size and stiffness with training

Shown to improve tendon size alongside training in healthy adults, with stiffness gains at 15 to 30 g a day, though certainty was very low.

The numbers and sources

Effect: Across 19 RCTs (768 healthy adults) training for 8 weeks or more, collagen peptides improved tendon morphology (SMD 0.67) [1]. In a systematic review of 8 RCTs (257 people, 246 of them men) with resistance or plyometric training for 3 to 15 weeks, tendon cross-sectional area grew more on collagen than placebo in 3 of 4 studies. Tendon stiffness rose more on collagen in the 2 studies using 15 to 30 g a day, and the studies using about 5 g a day showed within-group changes only [2].
Population: Healthy adults over 17 in exercise programs [1]; trained adults aged 18 to 52 [2].
Sources: [1], [2]
Caveats: GRADE certainty for tendon outcomes was very low [1]. Trials were small and short, and nearly all participants were men [2]. The second review did not pool results [2].

CollagenStrong

Joint pain and function

Shown to ease joint pain and improve joint function in recreational athletes and others, in reviews that did not pool results.

The numbers and sources

Effect: A systematic review of 15 RCTs, 12 in recreational athletes, reported the most consistent benefit for joint function and joint pain [3]. A systematic review of 36 RCTs of type I collagen hydrolysate found joint studies reported less pain, better clinical scores, more mobility and better ankle function [4].
Population: Recreational athletes, older adults and premenopausal women [3]; RCTs of type I collagen alone, with varied populations and health conditions [4].
Sources: [3], [4]
Caveats: Both reviews are narrative, without pooled effect sizes, and heterogeneity among trials limits generalization [4]. In a 2026 trial in 74 healthy people with joint discomfort during activity, 40 mg a day of undenatured type II collagen for 180 days improved joint scores, but so did placebo, and between-group differences appeared only in subgroups [5].

CuppingStrong

Chronic pain, compared with no treatment or usual care

Shown to ease chronic pain by a large amount against no treatment in the short term, and neck pain and fibromyalgia pain against no intervention or usual care.

Compared with no treatment [2]; no intervention [7]; usual care [6]

The numbers and sources

Effect: Against no treatment, cupping had a large short-term effect on pain (SMD -1.03, 95% CI -1.41 to -0.65) and a medium effect on disability (SMD -0.66) [2]. For neck pain, pain fell more than with no intervention (mean difference -2.42, 95% CI -3.98 to -0.86) and function improved [7]. In fibromyalgia, pain at day 18 was 12.4 points lower than with usual care (95% CI -18.9 to -5.9) [6].
Population: 1,172 people in 18 trials [2]. 18 trials of people with neck pain [7]. 141 patients with fibromyalgia [6].
Sources: [2], [7], [6]
Caveats: Without a sham, participants know they were treated. The same reviews and trials found no difference from sham cupping [2][6]. The neck pain reviewers rated the evidence low in quality [7].

Dry needlingStrong

Musculoskeletal pain, all comparisons pooled

Shown to ease musculoskeletal pain more than no treatment, sham needling and other therapies pooled together, from 72 hours to 24 weeks, in low-quality evidence.

Compared with no treatment, sham dry needling and other therapies, pooled together [3][4]

The numbers and sources

Effect: Across 42 trials, pain fell more than in the comparison groups within 72 hours (SMD -0.81, 95% CI -1.21 to -0.40), at 1 to 3 weeks (SMD -0.69, 95% CI -1.02 to -0.35), at 4 to 12 weeks (SMD -0.85, 95% CI -1.30 to -0.40) and at 13 to 24 weeks (SMD -0.81, 95% CI -1.64 to -0.03) [4]. An earlier review of 13 trials found less pain and higher pressure pain thresholds than control, sham or other treatment up to 12 weeks. At 6 to 12 months the difference in pain was not significant. Function improved against no treatment or sham, and did not differ from other physical therapy treatments [3].
Population: 42 trials of dry needling by physical therapists for musculoskeletal pain [4]. 13 trials of musculoskeletal pain conditions [3].
Sources: [3], [4]
Caveats: Both reviews pooled sham, no-treatment and other-treatment comparisons together, so the share of the effect beyond sham cannot be read from them. Heterogeneity downgraded the evidence to low quality at most time points [4], and the earlier review rated it very low to moderate [3]. The sham-controlled trials alone are in the next block.

GnRH drugs for fibroidsStrongStudied in women

Add-back hormones and bone during GnRH treatment

Shown to cut spine bone loss from about 5 to 6% to about 1 to 1.5% over a year when estrogen and progestin add-back accompanied elagolix or leuprolide.

Compared with the same GnRH drug without add-back [7][9][10][12]

The numbers and sources

Effect: After 12 months, lumbar spine bone density fell 4.8% on elagolix alone and 1.5% on elagolix with add-back (difference -3.3, 95% CI -4.1 to -2.5) [9]. More women lost bone density on elagolix alone than with estradiol and norethindrone acetate [7]. On leuprolide alone, spine density fell 3.2% at 24 weeks and 6.3% at 52 weeks, against 0.3% and 1.0% with norethindrone acetate 5 mg a day [10]. With GnRH agonists for fibroids, tibolone (SMD 0.36, 3 studies), raloxifene (SMD 1.01, 1 study) and estriol and ipriflavone (single studies) limited bone loss, while tibolone was linked to larger uterine volume and more bleeding [12].
Population: Women with fibroids in elagolix trials [7][9]; women with endometriosis in the leuprolide add-back trials, 29 to 115 per arm [10]; 622 women with fibroids in 12 add-back trials [12].
Sources: [9], [7], [10], [12]
Caveats: The leuprolide add-back data come from endometriosis trials, and the Lupron Depot label states it is not indicated with norethindrone acetate add-back for fibroid anemia [10]. Add-back trials followed women for up to 6 months, too short to measure fractures [12].

94 more

Hypnotherapy and clinical hypnosisStrong

Chronic pain

Shown to ease chronic pain moderately (Hedges' g -0.42) against control treatments, with larger effects after 8 or more sessions.

Compared with control interventions the source abstract does not describe [13]; standard care, and other psychological interventions [14]

The numbers and sources

Effect: In chronic musculoskeletal and nerve pain, hypnosis lowered pain intensity (Hedges' g -0.42 after treatment, -0.37 at short-term follow-up) and pain interference (-0.39). With 8 or more sessions the effect was g -0.555, and with fewer it was smaller and not significant [13]. An earlier meta-analysis found a moderate benefit over standard care, and a moderate benefit over other psychological treatments in people with pain other than headache [14].
Population: 530 adults in 9 trials [13]. Trials in chronic pain [14].
Sources: [13], [14]
Caveats: Heterogeneity across trials was large [14].

MeditationStrong

Chronic pain

Shown to ease chronic pain slightly in pooled trials rated low in quality. In back pain, an 8-week course matched CBT and beat usual care.

Compared with treatment as usual, passive controls, and education or support groups [8]; active control programs [3]; usual care and cognitive behavioral therapy [9]

The numbers and sources

Effect: Across 30 trials, pain fell more with mindfulness meditation than with controls (SMD 0.32, 95% CI 0.09 to 0.54), smaller in trials rated fair or higher in quality (SMD 0.19, 95% CI 0.03 to 0.34, 19 trials) [8]. Against active controls, pain improved by an effect size of 0.33 (95% CI 0.03 to 0.62) [3]. In 342 adults with chronic low back pain, clinically meaningful improvement in function at 26 weeks reached 60.5% with mindfulness-based stress reduction, 57.7% with CBT and 44.1% with usual care (relative risk against usual care 1.37, 95% CI 1.06 to 1.77, and against CBT 0.95) [9].
Population: 3,536 adults with chronic pain from migraine, headache, back pain, osteoarthritis or nerve pain in 38 trials [8]. 342 adults aged 20 to 70 with chronic low back pain [9].
Sources: [8], [3], [9]
Caveats: Heterogeneity was substantial (I² 77.6%) and evidence for pain was rated low quality [8]. The back pain course included gentle yoga as well as meditation [9].

Menopausal hormone therapyStrongStudied in women

Bone density

Shown to raise bone density in postmenopausal women, with total hip density up 3.7% over 3 years against almost none on placebo.

The numbers and sources

Effect: In the WHI, total hip bone density rose 3.7% over 3 years on estrogen plus progestin, against 0.14% on placebo [3]. Across 24 RCTs, lumbar bone density improved more on hormone therapy than control (SMD 1.52, 95% CI 1.33 to 1.71), and hip density improved with each regimen studied (SMDs 0.57 to 1.36) [2]. In a network meta-analysis of 16 RCTs, hormone therapy alone and with antiresorptive drugs raised lumbar spine density against control [6].
Population: 1,024 women with bone density measured in the WHI trial [3]; 5,089 women in 24 RCTs [2]; 2,861 postmenopausal women in 16 RCTs [6].
Sources: [3], [2], [6]
Caveats: Bone density is not fracture. Controls in the 24-trial review were not all placebo [2].

Muscle relaxantsStrong

Acute low back pain

Shown to lower acute low back pain by about 8 points on a 100-point scale within 2 weeks, with no gain in disability and more side effects.

The numbers and sources

Effect: In acute low back pain, non-benzodiazepine muscle relaxants lowered pain intensity at 2 weeks or less by 7.7 points on a 0 to 100 scale against control (95% CI 3.3 to 12.1), with no significant change in disability (-3.3, 95% CI -7.3 to 0.7). Adverse events were more frequent (RR 1.6, 95% CI 1.2 to 2.0) [1]. An earlier Cochrane review found non-benzodiazepine relaxants better than placebo for pain relief after 2 to 4 days (RR 0.80, 95% CI 0.71 to 0.89) and global efficacy (RR 0.49, 95% CI 0.25 to 0.95), with more adverse events (RR 1.50, 95% CI 1.14 to 1.98) [2]. For cyclobenzaprine, people were more likely to report improvement by day 14 than on placebo (OR 4.7, 95% CI 2.7 to 8.1), a number needed to treat of 2.7, with effect sizes of 0.38 to 0.58 that were largest in the first 4 days [3].
Population: 6,505 adults with non-specific low back pain in 31 pooled trials, from 49 reviewed [1]; 30 trials, 24 of them in acute low back pain [2]; randomized placebo-controlled trials of cyclobenzaprine in adults with back pain [3].
Sources: [1], [2], [3]
Caveats: The most recent review rated its evidence very low certainty and described the pain reduction as not clinically important, with most trials at high risk of bias [1]. A separate review found fair evidence for cyclobenzaprine and tizanidine in acute back or neck pain and very limited or inconsistent data for methocarbamol and baclofen [4].

NSAIDsStrong

Joint pain in knee or hip osteoarthritis

Shown to ease knee osteoarthritis pain and stiffness over placebo, including topical diclofenac. Pain relief was similar to exercise therapy.

The numbers and sources

Effect: Pain relief did not differ between exercise therapy and oral NSAIDs or paracetamol at about 4 weeks (SMD -0.12, 95% CrI -1.74 to 1.50), 8 weeks or 24 weeks. Function also did not differ at any time point [2]. Against placebo in knee osteoarthritis, etoricoxib improved WOMAC pain (MD -0.44, 95% CI -0.61 to -0.26), naproxen WOMAC function (MD -0.43, 95% CI -0.82 to -0.04) and diclofenac WOMAC stiffness (MD -0.40, 95% CI -0.67 to -0.13) [3]. Topical diclofenac gel, solution and patch each lowered knee pain and improved function against placebo. The patch gave the largest relief at 1 to 2 weeks (SMD -0.64, 95% CI -0.90 to -0.39), and skin and whole-body side effects stayed low [4].
Population: 152 RCTs with 17,431 people with knee or hip osteoarthritis [2]; 31 RCTs with 68,539 people with knee osteoarthritis comparing 16 treatments [3]; 12 RCTs of topical diclofenac in knee osteoarthritis [4].
Sources: [2], [3], [4]
Caveats: The comparison with exercise is mostly indirect, and NSAIDs were pooled with paracetamol. Few trials reported 8- and 24-week results [2]. The network estimates in [3] compare drugs indirectly. [4] covered up to 12 weeks.

Opioid pain relieversStrong

Knee and hip osteoarthritis pain

Shown to lower knee and hip osteoarthritis pain by small amounts against placebo, about 0.7 points of 10 for oxycodone-type opioids and 4% for tramadol.

The numbers and sources

Effect: In 22 trials, 10 of them of oral oxycodone, opioids lowered pain by 0.7 cm on a 10 cm scale against placebo (SMD -0.28, 95% CI -0.35 to -0.20), a number needed to treat of 10, and improved function by 0.6 units on a 0 to 10 scale. Effects were larger in trials of 4 weeks or less. Dropouts over adverse events were 6.4% against 1.7% [9]. In 22 trials of tramadol, pain improved 4% more than on placebo (95% CI 3% to 5%), and 15 of 100 against 10 of 100 improved by 20%. Physical function improved 4% more (95% CI 2% to 6%). The authors judged neither an important benefit [10].
Population: 8,275 people with knee or hip osteoarthritis in 22 trials [9]; 22 trials with 3,871 participants randomized to tramadol alone or with acetaminophen and 2,625 to placebo or an active control, mean duration 2 months [10].
Sources: [9], [10]
Caveats: Most tramadol trials were funded by industry, and 10 of 22 had incomplete outcome data [10]. The first review found signs of small-study bias [9].

Oral corticosteroidsStrong

Rheumatoid arthritis

Shown to ease joint tenderness and pain in rheumatoid arthritis more than placebo or NSAIDs, and to slow joint damage on X-ray over 1 to 2 years.

The numbers and sources

Effect: Low-dose prednisolone, at no more than 15 mg a day, improved joint tenderness against placebo by 12 tender joints (95% CI 6 to 18) and improved pain (SMD 1.75, 95% CI 0.87 to 2.64) and grip strength. Against NSAIDs it improved joint tenderness by 9 joints (95% CI 5 to 12) and pain (SMD 1.25, 95% CI 0.26 to 2.24) [4]. Across 15 studies, mostly in early disease and added to other disease-modifying drugs, glucocorticoids slowed X-ray progression (SMD 0.40, 95% CI 0.27 to 0.54), with a similar effect at 2 years (0.42, 95% CI 0.30 to 0.55) [5]. In a 2-year trial of prednisone 10 mg a day, clinical improvement over placebo was seen mainly in the first 6 months and X-ray progression was lower after month 6, with more osteoporotic fractures in the prednisone group [6].
Population: 320 people with rheumatoid arthritis in 10 short-term studies [4]; 1,414 people in 15 studies [5]; people with early active rheumatoid arthritis not yet on disease-modifying drugs [6].
Sources: [4], [5], [6]
Caveats: The short-term review covers small trials [4]. The mean cumulative dose in the X-ray studies was 2,300 mg of prednisone equivalent in the first year [5].

Osteopathy and osteopathic manipulative treatmentStrong

Low back pain, compared with a mix of sham, usual care, physiotherapy and no treatment

Shown to ease low back pain by about 13 points on a 100-point scale at 3 months, in pooled trials that mixed sham, usual care, physiotherapy and no-treatment comparisons.

Compared with sham treatment, usual medical care, physiotherapy and no treatment, pooled together [4]; control interventions the source abstract does not describe [5]; active treatment, placebo and no treatment [6]

The numbers and sources

Effect: At 3 months, pain fell 12.91 points more on a 100-point scale (95% CI -20.00 to -5.82) and function improved (SMD -0.36, 95% CI -0.58 to -0.14) in acute and chronic low back pain. In chronic low back pain alone, pain fell 14.93 points more (95% CI -25.18 to -4.68) [4]. A 2021 review of chronic low back pain found pain fell more than with control interventions (effect size -0.59, 95% CI -0.81 to -0.36) and function improved (-0.42) [5]. An earlier review found a smaller pain reduction (effect size -0.30, 95% CI -0.47 to -0.13) that held at short, intermediate and long-term follow-up [6].
Population: 15 trials, 10 in nonspecific low back pain [4]. 10 trials in chronic low back pain [5]. 6 trials with 8 comparisons [6].
Sources: [4], [5], [6]
Caveats: None of the trials in the 2021 review was judged at low risk of bias in every domain [5]. The comparisons could not be analyzed separately [4], so the pooled effect includes trials against no treatment, where patients knew they were untreated. The sham-only result is in the block above.

Physical therapy and exercise therapyStrong

Chronic low back pain, compared with no treatment, usual care or placebo

Shown to ease chronic low back pain about 15 points on a 100-point scale against no treatment, usual care or placebo, with a smaller gain in function.

Compared with no treatment, usual care or placebo [2]; no-exercise true control [3]

The numbers and sources

Effect: Exercise lowered pain by 15.2 points on a 100-point scale at earliest follow-up (95% CI -18.3 to -12.2), a difference the reviewers called clinically important, and functional limitation by 6.8 points (95% CI -8.3 to -5.3), short of their clinical threshold [2]. In a network meta-analysis against true control, Pilates had the largest pain effect (SMD -1.86), and resistance and stabilization or motor control exercise the largest function effects (SMD -1.14 and -1.13). Stretching and McKenzie exercise did not differ from true control for pain or function [3].
Population: 249 trials of adults with chronic low back pain, mean age 43.7, 59% women [2]. 5,578 patients in 89 trials [3].
Sources: [2], [3]
Caveats: Most trials were at risk of performance bias because exercise cannot be blinded, and heterogeneity and publication bias lowered certainty [2]. The network meta-analysis rated its evidence low quality [3].

Red light therapyStrong

Knee osteoarthritis pain

Shown to reduce knee osteoarthritis pain compared with placebo, by about 14 mm on a 100 mm scale in one meta-analysis. A later review rated certainty very low.

The numbers and sources

Effect: Across 22 placebo-controlled trials (1,063 participants), low-level laser reduced pain by 14.23 mm on a 100 mm visual analogue scale at the end of therapy (95% CI 7.31 to 21.14) and by 15.92 mm at follow-ups 1 to 12 weeks later (95% CI 6.47 to 25.37). At the doses advised by the World Association for Laser Therapy, the reduction was 18.71 mm. Disability was also reduced [6]. Across 10 placebo-controlled studies (542 participants), pain at rest was lower with photobiomodulation (-0.7, 95% CI -1.1 to -0.2), and the Timed Up and Go test did not differ [7].
Population: 1,063 people with knee osteoarthritis in 22 trials [6]. 542 people in 10 trials [7].
Sources: [6], [7]
Caveats: The later review judged every study at unclear to high risk of bias and the certainty of evidence very low [7]. The first review states that major knee osteoarthritis treatment guidelines do not endorse low-level laser [6]. These trials used clinic lasers applied to the joint, not consumer light panels.

Sodium levelStrong

Falls and fractures at low sodium

Linked to about twice the odds of falls and of fractures at low sodium in two meta-analyses of observational studies.

The numbers and sources

Effect: In a meta-analysis of 15 studies, hyponatremia was associated with more falls (OR 2.14, 95% CI 1.71 to 2.67). The odds were already higher when the cut-off for hyponatremia was 135 mmol/L (OR 1.26, 1.23 to 1.29). Hyponatremia was also associated with more fractures, particularly hip fractures (OR 2.00, 1.43 to 2.81) [12]. A second meta-analysis found more fractures with hyponatremia in studies that reported odds ratios (OR 1.99, 95% CI 1.50 to 2.63) and in studies that reported hazard ratios (HR 1.62, 1.28 to 2.05) [13].
Population: 51,879 patients in 15 studies, 2,329 of them hyponatremic [12]. 212,889 participants in 15 observational studies, 12 of them pooled [13].
Sources: [12], [13]
Caveats: Observational. The authors of one review write that it is unclear whether hyponatremia is directly associated with fracture or is a surrogate marker of other causes, and call for studies of fracture risk after hyponatremia is corrected [13].

Tai chi and qigongStrong

Falls in older adults

Shown to reduce the rate of falls in older adults by about 19% to 43% in pooled trials, and the number of people who fall by about 20%.

Compared with control interventions not thought to reduce falls [1]; other treatments [2]; control groups the source abstract does not describe [3]

The numbers and sources

Effect: In a Cochrane review, tai chi reduced the rate of falls by 19% (rate ratio 0.81, 95% CI 0.67 to 0.99, 7 trials, 2,655 participants, low certainty) and the number of people who fell by 20% (risk ratio 0.80, 95% CI 0.70 to 0.91, 8 trials, high certainty) [1]. Another meta-analysis found falls cut over the short term (incidence rate ratio 0.57, 95% CI 0.46 to 0.70) and less over the long term (0.87), with fewer injurious falls (0.50 short term). Time to first fall did not change (hazard ratio 0.98, 95% CI 0.69 to 1.37) [2]. A third found fewer fallers (risk ratio 0.80) and a lower fall rate (0.69, 95% CI 0.60 to 0.80), larger with more frequent practice [3].
Population: Community-dwelling adults aged 60 and over, mean age 76, 77% women [1]. Older and at-risk adults in 10 trials [2]. 3,824 older adults in 18 trials [3].
Sources: [1], [2], [3]
Caveats: Most trials had unclear or high risk of bias in at least one item [1]. Funnel plot asymmetry suggested publication bias [3]. The reviews share many trials.

Thunder god vineCautionStrong

Rheumatoid arthritis

Shown to improve joint swelling, stiffness and grip in adults with rheumatoid arthritis, matching or beating standard drugs, in trials mostly from China.

The numbers and sources

Effect: Across 14 RCTs, TwHF extracts improved grip strength, swollen joint count and morning stiffness against placebo (P < 0.001). Against DMARDs, grip strength, morning stiffness, CRP and tender joint count differed (P < 0.05), and ESR and swollen joint count did not [1]. Across 40 RCTs (3,092 patients), tripterygium glycosides added to DMARDs lowered morning stiffness, tender and swollen joint counts, pain scores, CRP, ESR and rheumatoid factor against DMARDs alone (all p < .001) [2]. A network meta-analysis of 21 trials (5,039 participants) ranked TwHF with methotrexate first and TwHF alone second for ACR20 response among the conventional DMARDs compared [3]. In 207 patients, ACR50 response at 24 weeks was 46.4% on methotrexate, 55.1% on TwHF (p 0.014) and 76.8% on both (p < 0.001) [4]. In 121 US patients, 65.0% on TwHF extract and 32.8% on sulfasalazine met ACR20 at 24 weeks (P 0.001) [5].
Population: Adults with active rheumatoid arthritis, mostly in China [1][2][3][4]. 121 patients at US centers [5].
Sources: [1], [2], [3], [4], [5]
Caveats: The methotrexate trial was open label, with blinded assessors [4]. In the sulfasalazine trial only 62 patients had outcome data at 24 weeks, and 62% and 41% stayed on treatment [5]. Most pooled trials were small and Chinese.

TurmericStrong

Knee osteoarthritis pain and function

Shown to lower knee pain and improve function more than placebo in adults with knee osteoarthritis, similar to NSAIDs, in short trials and low-quality reviews.

The numbers and sources

Effect: Across 16 RCTs (1,810 adults, up to 16 weeks), turmeric extracts reduced knee pain (SMD -0.82, 95% CI -1.17 to -0.47, I² 86.23%) and improved function (SMD -0.75, 95% CI -1.18 to -0.33) against placebo, with effects similar to NSAIDs. Higher BMI was linked to less improvement [1]. A second meta-analysis of 15 RCTs (1,621 participants) also reported lower VAS and WOMAC pain, function and stiffness scores than placebo and similar effects to NSAIDs [2]. A network meta-analysis of 17 RCTs found every turmeric preparation reduced WOMAC pain against placebo, by MD -3.17 (95% CI -5.50 to -0.83) for conventional curcuminoids and -2.47 (95% CI -3.27 to -1.67) for bioavailability-enhanced forms [3]. Another network meta-analysis of 20 RCTs (1,633 participants) found lower VAS pain against placebo only with modified Curcuma longa formulations (SMD -2.82, 95% CI -5.30 to -0.33) [4].
Population: Adults with knee osteoarthritis [1][2][3][4].
Sources: [1], [2], [3], [4], [5]
Caveats: Heterogeneity was high, trials were short, and risk of bias was moderate [1]. Imaging and biochemical markers did not differ [1]. Certainty was low in the 17-trial network analysis [3]. A review of 7 earlier systematic reviews rated their methods extremely low in quality, with heavy overlap of trials and 37 of 48 outcomes at extremely low evidence quality [5].

TurmericStrong

Rheumatoid arthritis

Shown to lower disease activity, joint counts and inflammation markers in adults with rheumatoid arthritis, in few small trials with very low certainty.

The numbers and sources

Effect: Across 6 publications (539 patients), curcumin lowered DAS28 (MD -1.20, 95% CI -1.85 to -0.55), ESR (MD -29.47), swollen joint count (MD -5.33) and tender joint count (MD -6.33) against controls [19]. Across 7 RCTs, DAS28 fell by 1.47 (95% CI -1.68 to -1.26), with lower rheumatoid factor, ESR and CRP [20]. Across 6 placebo-controlled trials (244 participants), ACR20 response, DAS28, ESR, CRP, pain and joint counts all improved, with very low certainty for most outcomes [21].
Population: Adults with rheumatoid arthritis [19][20][21].
Sources: [19], [20], [21]
Caveats: The trials were few and small, heterogeneity was substantial [20], and the certainty of evidence was very low [21].

Uric acid (serum urate)Strong

Gout

Linked to gout in a graded way: 15-year incidence ran from 1.1% below 6 mg/dL to 49% at 10 mg/dL or above.

The numbers and sources

Effect: In pooled individual data from four US cohorts, the 15-year cumulative incidence of gout was 1.1% (95% CI 0.9 to 1.4) at a baseline serum urate below 6 mg/dL and 49% (31 to 67) at 10 mg/dL or above. Against below 6 mg/dL, adjusted hazard ratios were 2.7 at 6.0 to 6.9 mg/dL, 6.6 at 7.0 to 7.9, 15 at 8.0 to 8.9, 30 at 9.0 to 9.9 and 64 at 10 or above [9]. In 2,046 initially healthy men followed for 14.9 years, the annual incidence of gouty arthritis was 4.9% at 9 mg/dL or more, 0.5% at 7.0 to 8.9 mg/dL and 0.1% below 7.0 mg/dL [10]. In a Mendelian randomization study, a genetically higher serum urate of 1 standard deviation went with an odds ratio for gout of 5.84 (95% CI 4.56 to 7.49) [11]. An umbrella review of 136 health outcomes found convincing evidence of a clear role for serum uric acid only for gout and kidney stones [12].
Population: 18,889 gout-free participants followed for a mean of 11.2 years [9]. 2,046 men in the Normative Aging Study [10]. 3,151 gout cases and 68,350 controls [11].
Sources: [9], [10], [11], [12]
Caveats: About half of the people at 10 mg/dL or above did not develop gout within 15 years, which the authors read as a role for other factors and for how long the level stays high [9].

Vitamin DStrong

Rheumatoid arthritis disease activity

Shown to lower disease activity modestly in adults with rheumatoid arthritis. Reviews disagree on CRP, ESR and pain.

The numbers and sources

Effect: Across 6 studies (438 patients), vitamin D lowered DAS28 (WMD -0.41, 95% CI -0.59 to -0.23), ESR (WMD -3.40) and tender joint count (WMD -1.44) against control [6]. Across 11 RCTs, it lowered pain scores (WMD -1.30) and DAS28-CRP (WMD -0.58, 95% CI -0.86 to -0.31), while CRP, ESR and HAQ did not change [7]. A third meta-analysis of 11 studies found lower DAS28 (WMD -0.83, 95% CI -1.38 to -0.28), CRP and ESR, with no change in HAQ or pain scores [8].
Population: Adults with rheumatoid arthritis in RCTs [6][7][8].
Sources: [6], [7], [8]
Caveats: The reviews disagree on CRP, ESR and pain, and heterogeneity was high [7].

YogaStrong

Balance and mobility in older adults

Shown to improve balance by a small to moderate amount in adults aged 60 and over, in two meta-analyses. Whether this prevents falls was not tested.

Compared with control groups the source abstract does not describe [13]; inactive controls such as a waiting list or education booklets [7]

The numbers and sources

Effect: Yoga improved balance performance (g = 0.40, 95% CI 0.15 to 0.65, 6 trials) and physical mobility (g = 0.50, 95% CI 0.06 to 0.95, 3 trials) [13]. In older adults without a specific condition, balance improved against inactive controls (effect size 0.7) [7].
Population: 307 people aged 60 and over in 6 trials [13]. Older adults in 22 trials [7].
Sources: [13], [7]
Caveats: The reviewers state that whether balance gains translate into fewer falls is unknown [13].

AcupunctureModerate

Chronic low back pain, compared with no treatment or usual care

Shown to ease chronic low back pain by about 20 points on a 100-point scale against no treatment. Against usual care the pain difference was smaller, about 10 points.

Compared with no treatment [4]; usual care [4]

The numbers and sources

Effect: Against no treatment, pain fell 20.32 points more on a 0 to 100 scale (95% CI -24.50 to -16.14, 4 trials, 366 participants) and back function improved (SMD -0.53, 95% CI -0.73 to -0.34, 5 trials, 2,960 participants), at moderate certainty. Against usual care, pain fell 10.26 points more (95% CI -17.11 to -3.40, 5 trials, 1,054 participants), which the reviewers did not count as clinically important, and function improved (SMD -0.47) [4].
Population: Adults with chronic nonspecific low back pain in 33 trials [4].
Sources: [4]
Caveats: Neither patients nor acupuncturists were blinded in these comparisons, so expectation is part of the effect. The effect was measured right after treatment [4].

AcupunctureModerate

Neck pain

Shown to ease neck pain slightly more than sham acupuncture in the short term, and more than a waiting list. The effect did not appear to last.

Compared with sham acupuncture [5]; inactive treatment [5]; waiting list [5]

The numbers and sources

Effect: For mechanical neck pain, pooled sham-controlled trials favored acupuncture in the short term (SMD -0.23, P = 0.0006). Acupuncture also eased pain more than inactive treatment, and pain and disability more than a waiting list, at short-term follow-up. The reviewers write that the effect does not seem sustainable over the long term [5].
Population: 27 trials of adults with whiplash, myofascial, arthritic, nonspecific, radicular or mechanical neck pain [5].
Sources: [5]
Caveats: Evidence was rated moderate in quality. Few trials were large, and repeated courses were not studied [5].

AllopurinolModerateVery common

Lowering uric acid in gout

Shown to lower blood uric acid, with 25 of 26 people on 300 mg a day reaching the target within 30 days against none of 25 on placebo.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: In the one placebo-controlled trial in a Cochrane review, 25 of 26 people on allopurinol 300 mg a day reached the target serum urate over 30 days against 0 of 25 on placebo [1]. At a fixed 300 mg a day for 52 weeks, 21% had urate below 6.0 mg/dL at the last three monthly measurements [2]. When the dose was raised step by step, 69% had urate below 6 mg/dL at 12 months against 32% of those left on their starting dose, with mean changes of -1.5 and -0.34 mg/dL [3]. In a 72-week trial that raised doses toward a urate target, 80% reached it on allopurinol or febuxostat with no difference between them [4].
Population: 4,531 participants in 11 trials, 2 of them placebo-controlled [1]; 760 people with gout and urate of at least 8.0 mg/dL [2]; 183 people with gout already on allopurinol, 52% with reduced kidney function [3]; 940 people with gout, at least a third with stage 3 chronic kidney disease [4].
Sources: [1], [2], [3], [4]
Caveats: The placebo comparison rests on one trial of 57 people [1]. The share reaching target depended on whether the dose was fixed or raised [2][3].

Biofeedback and neurofeedbackModerate

Chronic back pain

Shown to reduce chronic back pain intensity with a moderate effect (Hedges' g 0.60) that held over about 8 months, alone or added to other treatment.

Compared with control conditions the source abstract does not describe, with before-and-after studies included [10]

The numbers and sources

Effect: Biofeedback reduced pain intensity (g 0.60, 95% CI 0.44 to 0.76), and the effect held at an average 8-month follow-up (g 0.62). It also reduced depression (g 0.40), disability (g 0.49) and muscle tension (g 0.44). Effects held when only controlled studies were analyzed [10].
Population: 1,062 patients in 21 studies [10].
Sources: [10]
Caveats: The analysis included uncontrolled studies, and the abstract does not describe the controls [10].

BisphosphonatesModerateStudied in women

Fractures in lower-risk women and women with osteopenia

Shown to lower spine and other fractures in lower-risk postmenopausal women, with 15 women with osteopenia treated with zoledronate for 6 years per fracture prevented.

Compared with placebo or no treatment [1]; placebo [2][6]

The numbers and sources

Effect: In lower-risk women, alendronate lowered clinical vertebral fractures (RR 0.45, 95% CI 0.25 to 0.84, 1.4% fewer) and non-vertebral fractures (RR 0.83, 1.6% fewer), with no clear difference in hip fractures (RR 0.76, 95% CI 0.43 to 1.32) or wrist fractures [1]. For risedronate in lower-risk women, effects on fractures were uncertain [2]. In women aged 65 or older with osteopenia, a fragility fracture occurred in 122 on zoledronate and 190 on placebo over 6 years (HR 0.63, 95% CI 0.50 to 0.79), a number needed to treat of 15 [6].
Population: 10,057 lower-risk women in 16 alendronate trials of 1 to 5 years [1]; 989 women in 4 risedronate trials [2]; 2,000 women aged 65 or older with osteopenia, median 10-year hip fracture risk 2.3% [6].
Sources: [1], [2], [6]
Caveats: Low-certainty evidence in [1]. The osteopenia result rests on one trial [6].

BisphosphonatesModerateVery common

Paget's disease of bone

Shown to relieve bone pain in Paget's disease more often than placebo, 31% against 9%, with one zoledronic acid infusion normalizing alkaline phosphatase in 89% at 6 months.

Compared with placebo and other bisphosphonates [12]; risedronate [13]

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Bone pain disappeared in 31% on a bisphosphonate and 9% on placebo (RR 3.42, 95% CI 1.31 to 8.90), a number needed to treat of 5. Fractures did not clearly differ. Zoledronate relieved pain more often than pamidronate or risedronate in single trials [12]. In two 6-month trials, one 5 mg infusion of zoledronic acid normalized serum alkaline phosphatase in 89% against 58% on risedronate 30 mg daily for 2 months, and a therapeutic response was seen in 96% against 74% [13].
Population: 3,168 people in 20 trials, mean age 66 to 74, 51% to 74% men, mostly with raised alkaline phosphatase [12]; 349 adults with moderate to severe Paget's disease, mean age 70 [13].
Sources: [12], [13]
Caveats: Moderate-quality evidence for pain, and mean follow-up was 6 months [12]. Treating to normalize alkaline phosphatase did not improve pain or quality of life over treating symptoms [12].

CalciumModerate

Bone mineral density

Shown to raise bone density by 0.7% to 1.8% in people over 50, a gain the authors judged unlikely to meaningfully cut fractures.

The numbers and sources

Effect: Calcium supplements raised bone density by 0.7 to 1.8% at five skeletal sites. The gain at later time points was similar to the gain at one year. Calcium from food raised it by similar amounts at most sites [6]. Hip density was 1.06% higher with calcium plus vitamin D than placebo in postmenopausal women [7]. In postmenopausal women with osteoporosis, calcium plus vitamin D raised pelvic density slightly (SMD 0.20, 95% CI 0.05 to 0.35), with no significant change at the lumbar spine, femoral neck or total hip [2].
Population: 59 RCTs in people over 50, 51 of them of supplements (12,257 people) [6]; 36,282 postmenopausal women aged 50 to 79 [7]; 11 RCTs in women with osteoporosis [2].
Sources: [6], [7], [2]
Caveats: The review authors write that gains of this size are unlikely to lead to a clinically meaningful drop in fractures [6].

Chiropractic and spinal manipulationModerate

Acute low back pain

Shown to ease acute low back pain by about 10 points on a 100-point scale and improve function modestly for up to 6 weeks, against sham or other treatments.

Compared with sham or alternative treatments [6]

The numbers and sources

Effect: Across 15 trials (1,711 patients), pain improved by 9.95 points on a 100-mm scale (95% CI -15.6 to -4.3). Across 12 trials (1,381 patients), function improved (effect size -0.39, 95% CI -0.71 to -0.07). Heterogeneity was not explained by who gave the manipulation or by type of manipulation [6].
Population: Adults with acute low back pain treated in ambulatory settings, 26 trials [6].
Sources: [6]
Caveats: The abstract pools sham and active comparisons together and calls heterogeneity large. Outcomes ran only to 6 weeks [6].

Chiropractic and spinal manipulationModerate

Neck pain

Shown to ease acute and subacute neck pain more than inactive controls in the short term with thoracic manipulation. Neck manipulation and mobilization gave similar results.

Compared with inactive controls [7]; mobilization [7]; certain medications [7]

The numbers and sources

Effect: Thoracic manipulation against inactive control improved pain at short-term follow-up (SMD -1.26, 95% CI -1.86 to -0.66, 5 trials, 346 participants) and function (SMD -1.40, 4 trials), at moderate quality. A single neck manipulation relieved pain immediately but not at short-term follow-up, at low to very low quality. Multiple sessions of neck manipulation produced similar results to mobilization, and better pain and function than certain medications in one trial of 182 people [7].
Population: 51 trials with 2,920 participants with neck pain [7].
Sources: [7]
Caveats: A funnel plot suggested publication bias for the thoracic manipulation results [7].

Chiropractic and spinal manipulationModerate

Migraine, compared with other control conditions

Shown to reduce migraine days by a small amount in a meta-analysis of 5 trials, a result the authors called preliminary.

Compared with control conditions the source abstract does not describe [8]

The numbers and sources

Effect: Spinal manipulation reduced migraine days (Hedges' g -0.35, 95% CI -0.53 to -0.16) and migraine pain or intensity. One trial in chronic migraine was left out of the pooled analysis because it made heterogeneity high [8].
Population: 677 people in 6 trials, with treatment lasting 2 to 6 months [8].
Sources: [8]
Caveats: Several trials were at high or unclear risk of bias for blinding, compliance and missing data [8].

Cognitive behavioral therapyModerate

Chronic pain

Shown to ease chronic pain, disability and distress slightly more than usual treatment. Against other active treatments the benefit was very small and did not last.

Compared with treatment as usual [15]; active control treatments [15]

The numbers and sources

Effect: Against treatment as usual, CBT lowered pain (SMD -0.22, 95% CI -0.33 to -0.10, 29 trials), disability (SMD -0.32) and distress (SMD -0.34) at treatment end, effects largely maintained at follow-up. Against active controls, benefits were very small for pain (SMD -0.09, 95% CI -0.17 to -0.01, 23 trials), disability (SMD -0.12) and distress (SMD -0.09), and not maintained at follow-up [15].
Population: 59 CBT trials among 75 studies with 9,401 adults, mostly with fibromyalgia, chronic low back pain, rheumatoid arthritis or mixed chronic pain [15].
Sources: [15]
Caveats: Evidence quality ranged from moderate to low, and most trials were at high or unclear risk of bias. Headache was excluded [15].

CollagenModerate

Osteoarthritis symptoms

Shown to lower total osteoarthritis scores by about 8 points and lower pain ratings, though the WOMAC pain subscore did not change.

The numbers and sources

Effect: In placebo-controlled trials, collagen lowered total WOMAC score by 8.00 points (95% CI -13.04 to -2.95) and VAS pain by 16.57 points (95% CI -26.24 to -6.89). Within WOMAC, stiffness fell (-0.41) but the pain (-0.22, 95% CI -1.58 to 1.13) and function subscores did not change [9].
Population: Adults with osteoarthritis [9].
Sources: [9]
Caveats: The pain subscore and the separate VAS pain score point in different directions. The abstract does not report the number of trials or their funding.

DHEAModerate

Bone density

Shown to raise hip bone density in older women, with no similar effect in men in pooled trials. No trial counted fractures.

The numbers and sources

Effect: In women, hip bone density rose more on DHEA than placebo, reported as an SMD of -0.5 for placebo against DHEA (95% CI -0.95 to -0.04). For the trochanter the SMD was -0.55 (95% CI -1.10 to 0.00, p = .05). No similar effect was seen in men [9]. In a 2-year trial, women on DHEA gained density at the ultradistal radius and men at the femoral neck [3]. In elderly men, no effect on bone health was found [2].
Population: Healthy older adults in placebo-controlled RCTs [9]; 144 elderly adults [3]; 1,353 elderly men [2].
Sources: [2], [3], [9]
Caveats: Results differ by sex and site. No sourced trial counted fractures.

Direct oral anticoagulants (DOACs)Moderate

Preventing clots after hip or knee replacement

Shown to prevent symptomatic clots after hip or knee replacement at least as well as enoxaparin injections, with rivaroxaban lower on clots and higher on bleeding.

The numbers and sources

Effect: Against enoxaparin, the risk of symptomatic venous thromboembolism was lower with rivaroxaban (RR 0.48, 95% CI 0.31 to 0.75) and similar with dabigatran (RR 0.71, 95% CI 0.23 to 2.12) and apixaban (RR 0.82, 95% CI 0.41 to 1.64). Clinically relevant bleeding was higher with rivaroxaban (RR 1.25, 95% CI 1.05 to 1.49), similar with dabigatran (RR 1.12, 95% CI 0.94 to 1.35) and lower with apixaban (RR 0.82, 95% CI 0.69 to 0.98) [12].
Population: 38,747 patients in 16 trials [12].
Sources: [12]
Caveats: The treatments did not differ on the net clinical endpoint. Edoxaban was not included [12].

Dry needlingModerate

Neck pain

Shown to ease neck pain in the short term against sham, placebo, waiting list or other needling pooled together. Neck function did not differ from physical or manual therapy.

Compared with sham, placebo, waiting list or another form of dry needling, pooled together [7]; manual therapy [7]; other physical therapy interventions [7]; sham dry needling, physical therapy or no treatment [8]

The numbers and sources

Effect: Against sham, placebo, waiting list or another form of dry needling, pain was lower immediately after treatment (mean difference -1.53 points, 95% CI -2.29 to -0.76) and in the short term (-2.31 points, 95% CI -3.64 to -0.99), and pain-related disability was lower in the short term (SMD -0.87). Against manual therapy, short-term pain was 0.51 points lower (95% CI -0.95 to -0.06) and disability did not differ. Against other physical therapy, no outcome differed. Nothing differed at mid-term, and neck range of motion did not change against any comparison [7]. In chronic nonspecific neck pain, function did not differ from control groups (mean difference 3.51, p = 0.14) and pain relief was borderline higher (mean difference 0.94, p = 0.05) [8].
Population: 28 trials of people with neck pain and trigger points [7]. 5 trials of chronic nonspecific neck pain [8].
Sources: [7], [8]
Caveats: Sham and waiting-list comparisons were pooled together, so the effect beyond sham is not separable. No trial could blind the therapist [7]. Evidence was rated low to moderate [7].

Dry needlingModerate

Low back pain

Shown to ease low back pain more than other treatments right after treatment in 11 trials. Effects on disability were smaller and effects at later follow-up unclear.

Compared with other treatments the review abstract does not list [9]; laser, physical therapy and other combined treatments [6]

The numbers and sources

Effect: Across 11 trials, dry needling of trigger points lowered pain intensity (SMD -1.06, 95% CI -1.77 to -0.36) and functional disability (SMD -0.76, 95% CI -1.46 to -0.06) more than other treatments after the intervention. Dry needling combined with other treatments lowered pain more than dry needling alone (SMD 0.83, 95% CI 0.55 to 1.11) [9]. A second review found results against laser, physical therapy and other combined treatments mixed, and called the evidence not robust enough for a firm conclusion [6].
Population: 802 people with low back pain and trigger points in 11 trials [9]. 16 trials of low back pain [6].
Sources: [9], [6]
Caveats: The reviewers call the effect on disability and at follow-up unclear [9]. Most trials were at high or unclear risk of bias [6].

Dry needlingModerate

Shoulder pain

Shown to ease nontraumatic shoulder pain by a small amount, about half a point, and disability by a larger amount, in the short term only.

Compared with comparison groups the review abstract does not describe [10]

The numbers and sources

Effect: Trigger point dry needling, alone or added to other treatment, lowered shoulder pain (mean difference -0.49 points, 95% CI -0.84 to -0.13, SMD -0.25) and pain-related disability (mean difference -9.99 points, 95% CI -15.97 to -4.01, SMD -1.14) more than the comparison groups. The effect on pain was found only in the short term [10].
Population: 6 trials of nontraumatic shoulder pain [10].
Sources: [10]
Caveats: Evidence was moderate in quality for pain and low for disability, downgraded for heterogeneity [10].

Dry needlingModerate

Tennis elbow

Shown to ease tennis elbow pain and disability with large effects in the short term, and grip strength with a small effect.

Compared with comparison groups the review abstract does not describe [15]

The numbers and sources

Effect: Dry needling, alone or with other treatment, lowered pain (SMD -1.13, 95% CI -1.64 to -0.62) and related disability (SMD -2.17, 95% CI -3.34 to -1.01), and raised pressure pain thresholds (SMD 0.98) and grip strength (SMD 0.48, 95% CI 0.16 to 0.81). The largest effects were in the short term [15].
Population: 320 people with lateral epicondylalgia in 7 trials [15].
Sources: [15]
Caveats: Evidence was low to moderate, downgraded for heterogeneity [15].

EstradiolModerate

Fractures in older women and men

Linked to more fractures at low levels: 2.5 times the hip fracture risk in older women with undetectable estradiol, and more fractures in older men below 16 pg/mL.

The numbers and sources

Effect: Among women aged 65 or older, those with undetectable estradiol (below 5 pg/mL) had a relative risk of 2.5 for hip fracture (95% CI 1.4 to 4.6) and 2.5 for vertebral fracture (1.4 to 4.2) against women with detectable levels [8]. In older Swedish men measured by mass spectrometry, each standard deviation lower estradiol had a hazard ratio for fracture of 1.34 (1.22 to 1.49). The relation was nonlinear and strong below 16 pg/mL [9].
Population: 133 women with later hip fracture and 138 with later vertebral fracture, compared with randomly selected women from the same cohort, estrogen users excluded [8]. 2,639 men with a mean age of 75, followed for 3.3 years on average [9].
Sources: [8], [9]
Caveats: Observational. The women's results were adjusted for age and weight only [8]. Many routine assays cannot quantify levels this low [4].

GingerModerate

Osteoarthritis pain

Shown to reduce osteoarthritis pain and disability modestly in adults over placebo.

The numbers and sources

Effect: In 5 placebo-controlled trials (593 patients), ginger reduced pain (SMD -0.30, 95% CI -0.50 to -0.09, I² 27%) and disability (SMD -0.22, 95% CI -0.39 to -0.04) [6].
Population: Adults with osteoarthritis [6].
Sources: [6]
Caveats: The authors rated the evidence moderate quality because trials were small and lacked proper intention-to-treat analysis [6].

Glucosamine and chondroitinModerateVery common

Chondroitin and pain in shorter trials

Shown to lower pain about 10% in trials under 6 months in people with osteoarthritis, mostly of the knee. Reviews disagree on whether this is meaningful.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Across 43 RCTs, chondroitin lowered pain by an absolute 10% (95% CI 6 to 15%) in studies under 6 months, with I² 70% and low-quality evidence. A 20% reduction in knee pain was reached by 53 per 100 on chondroitin and 47 per 100 on placebo (RR 1.12, 95% CI 1.01 to 1.24, 2 trials, high-quality evidence) [4].
Population: 4,962 participants on chondroitin and 4,148 controls, mostly knee osteoarthritis, 1 month to 3 years [4].
Sources: [4]
Caveats: This review and the network meta-analysis disagree on whether chondroitin's pain effect is clinically meaningful [1][4]. High heterogeneity was not explained by bias or sample size [4].

Heat therapyModerate

Acute low back pain, compared with an oral placebo

Shown to ease acute and subacute low back pain slightly against an oral placebo after 5 days, by about 1 point on a 0 to 5 scale, in two trials.

Compared with oral placebo [7]; a comparison the abstract does not describe, for a heated blanket [7]

The numbers and sources

Effect: In two trials of 258 people, a heat wrap reduced pain after five days against oral placebo (WMD 1.06, 95% CI 0.68 to 1.45, on a 0 to 5 scale). In one trial of 90 people, a heated blanket reduced acute pain immediately after use (WMD -32.20, 95% CI -38.69 to -25.71, on a 0 to 100 scale). There was not enough evidence to judge cold, and conflicting evidence on heat against cold [7].
Population: 1,117 people in 9 trials, most with acute or subacute low back pain [7].
Sources: [7]
Caveats: The reviewers call the evidence moderate for a small short-term effect, from few trials. No pooled trial covered chronic low back pain [7].

Horny goat weedModerate

Bone density after menopause

Shown to keep bone density steady over 2 years in postmenopausal women with low bone density, on a blend with soy isoflavones, while it fell on placebo.

The numbers and sources

Effect: In 100 late postmenopausal women with low bone density, a daily blend of 60 mg icariin, 15 mg daidzein and 3 mg genistein held bone density steady over 24 months. With placebo, lumbar spine density fell 2.4% and femoral neck density 1.8%. The groups differed at the lumbar spine at 12 and 24 months and at the femoral neck at 24 months. Both groups took 300 mg of calcium a day, and endometrial thickness showed no detectable hyperplasia [1]. A meta-analysis of 12 RCTs with 1,017 patients found higher bone density when Epimedium was added to conventional osteoporosis drugs (p = 0.03) or used in place of them (p = 0.009) [2]. In 58 postmenopausal women, 740 mg a day of Epimedium prenylflavonoids for 6 weeks raised the bone formation marker BSAP [3]. A 2025 meta-analysis of 10 RCTs (890 patients with primary osteoporosis) found higher lumbar spine bone density with Epimedium as an add-on or alternative to conventional drugs (SMD 1.15, 95% CI 0.61 to 1.70) [4].
Population: 100 postmenopausal women [1]; patients with osteoporosis in 12 RCTs [2] and 10 RCTs [4]; healthy postmenopausal women [3].
Sources: [1], [2], [3], [4]
Caveats: The blend in [1] also contained soy isoflavones. The trials in [2] and [4] compared Epimedium with conventional drugs, not placebo, and the abstract of [2] gives p-values without effect sizes. Review [4] searched Chinese databases.

Massage therapyModerate

Pain across conditions

Shown to ease pain more than no treatment, by a large margin, and more than sham or active treatments, by smaller margins, in a meta-analysis of 67 trials.

Compared with sham treatments [2]; no treatment [2]; active treatments [2]

The numbers and sources

Effect: Massage lowered pain more than sham (SMD -0.44), no treatment (SMD -1.14) and active comparators (SMD -0.26) [2]. A map of 49 systematic reviews, 32 rated high quality, found their conclusions gave low strength of evidence for potential benefits in labor, shoulder, neck, low back, cancer, arthritis, postoperative and muscle soreness pain [3].
Population: 67 randomized trials of people with pain in the general population, 60 rated high quality [2]. 49 systematic reviews [3].
Sources: [2], [3]
Caveats: The evidence map notes few large, rigorous trials and poor description of the massage given, so reviews could not say which type helps which pain [3]. Participants cannot be blinded to massage.

Massage therapyModerate

Low back pain

Shown to ease subacute and chronic low back pain more than inactive controls in the short term, with no lasting difference. Against active treatments the pain difference was small.

Compared with inactive controls [4]; active treatments [4]; usual care [5]

The numbers and sources

Effect: For subacute and chronic low back pain, massage lowered pain more than inactive controls in the short term (SMD -0.75, 95% CI -0.90 to -0.60, 7 trials, 761 participants) and improved function (SMD -0.72), with no difference in the long term. Against active controls, pain was lower in the short term (SMD -0.37, 12 trials, 964 participants) and long term (SMD -0.40), and function did not differ [4]. In 401 adults, Roland disability scores at 10 weeks were 2.9 points lower with relaxation massage and 2.5 points lower with structural massage than with usual care [5].
Population: 25 trials with 3,096 participants, almost all with subacute or chronic low back pain [4]. 401 adults aged 20 to 65 with chronic low back pain in Seattle [5].
Sources: [4], [5]
Caveats: The reviewers rated the evidence low to very low and wrote that they have very little confidence in massage for low back pain [4]. The trial did not blind participants to massage versus usual care [5].

Menopausal hormone therapyModerateStudied in women

Fractures

Shown to lower fractures in postmenopausal women, 8.6% against 11.1% on placebo, with fewer hip fractures. Overall benefits and harms showed no net benefit.

The numbers and sources

Effect: In the WHI estrogen plus progestin trial, 8.6% of women on treatment and 11.1% on placebo had a fracture (HR 0.76, 95% CI 0.69 to 0.83). The effect did not differ by age, weight, fall history or baseline bone density [3]. Hip fracture HR was 0.66 (95% CI 0.45 to 0.98), 5 fewer hip fractures per 10,000 person-years [4]. The estrogen-alone trial showed similar results for other outcomes, including fewer hip fractures [5].
Population: 16,608 postmenopausal women aged 50 to 79 with a uterus, followed a mean of 5.6 years [3]; 10,739 women after hysterectomy in the estrogen-alone trial [5].
Sources: [3], [4], [5]
Caveats: Even in women at the highest fracture risk, the global index of benefits and harms showed no net benefit (HR 1.03, 95% CI 0.88 to 1.24) [3]. In a network meta-analysis, fracture events were sparse and often not prespecified outcomes [6].

MSMModerate

Knee osteoarthritis pain and function

Shown to improve knee osteoarthritis physical function scores over placebo in adults, with pain results mixed and gains called small.

The numbers and sources

Effect: On 6 g a day for 12 weeks, WOMAC pain and physical function improved against placebo (P < 0.05). Stiffness and total symptom scores did not change [1]. On 3.4 g a day for 12 weeks, WOMAC physical function (14.6 mm, P = 0.04) and total score (15.0 mm, P = 0.03) improved. WOMAC pain did not reach significance (P = 0.08) [2]. On 1.5 g a day, the pain index fell from 1.53 to 0.74, and more with glucosamine added [3]. A 2008 review called the MSM trials positive but not definitive [4].
Population: 50 adults aged 40 to 76 [1]; 49 adults aged 45 to 90 [2]; 118 people with mild to moderate knee osteoarthritis [3]; 168 people in 2 MSM trials [4].
Sources: [1], [2], [3], [4]
Caveats: The authors of [2] call the improvements small and of uncertain clinical significance. The review raised concerns about dose and treatment length [4].

NSAIDsModerate

Low back pain

Shown to lower acute low back pain slightly with diclofenac, with other NSAIDs unconfirmed or showing no benefit.

The numbers and sources

Effect: Against placebo, diclofenac lowered pain intensity in acute low back pain. Ibuprofen in acute low back pain without nerve root pain and naproxen in acute low back pain with nerve root pain showed results the authors rated as needing confirmation. Celecoxib, etoricoxib and ketoprofen showed no evidence of benefit. The authors judged every effect small [5].
Population: 11 RCTs with 3,784 people, 1,642 with acute and 2,142 with chronic low back pain [5].
Sources: [5]
Caveats: 5 of the 11 trials were at high risk of bias for pain, and more than half used the NSAID as a comparator for another treatment [5].

Oral corticosteroidsModerate

Sciatica and low back pain

Shown to improve function slightly in sciatica from a herniated disk, with pain about half a point of 10 lower and no change in surgery rates.

The numbers and sources

Effect: In adults with sciatica from a herniated disk, a 15-day prednisone course improved disability at 3 weeks by 6.4 points more than placebo on a 100-point scale (95% CI 1.9 to 10.9) and by 7.4 points at 52 weeks. Leg pain did not differ at 3 weeks (0.3 points of 10, 95% CI -0.4 to 1.0) or 52 weeks, and surgery rates did not differ [9]. A Cochrane review of 13 trials found systemic corticosteroids probably lower radicular pain slightly in the short term (0.56 points of 10, 95% CI 0.04 to 1.08) and raise the likelihood of improved function (RR 1.52, 95% CI 1.22 to 1.91), with no difference in surgery (RR 1.00, 95% CI 0.68 to 1.47). For low back pain without sciatica and for spinal stenosis, the limited evidence showed no clear benefit [10].
Population: 269 adults with radicular pain for 3 months or less and a herniated disk on MRI, 181 on prednisone and 88 on placebo [9]; 1,047 participants in 13 trials [10].
Sources: [9], [10]
Caveats: The long-term function result rests on a single trial [10]. At least one adverse event was reported by 49.2% on prednisone and 23.9% on placebo at 3 weeks [9].

Ozone therapyModerate

Herniated lumbar disc, injection into the disc

Shown in pooled trials to give higher treatment success than steroid injections and more pain relief than medications at 6 months or more, with results similar to microdiscectomy.

Compared with steroid injections [5]; conventional medications [5]; microdiscectomy [5]

The numbers and sources

Effect: At 6 months, treatment success was higher than with steroid injections (OR 3.95, 95% CI 2.44 to 6.39). Pain fell more than with conventional medications at 6 months (SMD 1.65, 95% CI 1.08 to 2.22) and 12 months (SMD 1.52). At 18 months, pain did not differ from microdiscectomy (SMD -0.05, 95% CI -0.67 to 0.57) [5].
Population: People with herniated lumbar discs in trials searched to January 2024 [5].
Sources: [5]
Caveats: None of these comparisons was against a sham injection, so participants knew which treatment they had. A separate 2025 meta-analysis on disc injections of ozone has been retracted and is not used here.

Parathyroid hormone (PTH)Moderate

Vitamin D status

Linked to higher PTH at lower vitamin D: 40% of people with 25-hydroxyvitamin D under 20 ng/mL had PTH above 65 pg/mL in 312,962 paired results.

The numbers and sources

Effect: In paired PTH and 25-hydroxyvitamin D results from one reference laboratory, 40% of people with 25-hydroxyvitamin D below 20 ng/mL and 51% of those below 10 ng/mL had PTH above 65 pg/mL. Median PTH fell smoothly as vitamin D rose, with no level above which it stopped falling, and the relation depended strongly on age [8]. In adults aged 60 to 79, the upper limit of PTH was 64 pg/mL in the whole group and 46 in those with 25-hydroxyvitamin D above 30 nmol/L [5]. With a third-generation assay, taking the reference group only from people with 25-hydroxyvitamin D of 30 ng/mL or above lowered the upper limit by 22.4% [9].
Population: 312,962 paired results measured from July 2010 to June 2011 [8]. 280 adults aged 60 to 79 [5]. 898 adults aged 18 to 89 with normal body mass index and kidney function, 183 of them vitamin D replete [9].
Sources: [8], [5], [9], [7], [6]
Caveats: Cross-sectional. The large series is of people whose doctors ordered both tests, not a reference population. Sources disagree on whether the reference range is to be lowered to account for vitamin D: two studies set upper limits near 46 pg/mL [5][7], and a third concluded the limit stays at 65 [6].

Physical therapy and exercise therapyModerate

Knee osteoarthritis, compared with no exercise

Shown to ease knee osteoarthritis pain about 12 points on a 100-point scale and improve function about 10 points against no exercise, right after treatment.

Compared with non-exercise or no-treatment control groups [4]

The numbers and sources

Effect: Exercise reduced pain (SMD -0.49, 44 trials, 3,537 participants), equal to 12 points on a 0 to 100 scale, and improved physical function by 10 points (SMD -0.52), both immediately after treatment. Quality of life improved by 4 points. Two to six months after treatment ended, pain was 6 points lower and function 3 points better [4].
Population: 54 trials of people with knee osteoarthritis [4].
Sources: [4]
Caveats: Only 4 trials blinded participants, and pain and function were self-reported [4].

Physical therapy and exercise therapyModerate

Hip osteoarthritis, compared with no exercise

Shown to ease hip osteoarthritis pain and improve function about 8 and 7 points on a 100-point scale against no exercise, sustained 3 to 6 months later.

Compared with non-exercise control groups [5]

The numbers and sources

Effect: Exercise reduced pain (SMD -0.38, 9 trials, 549 participants), equal to 8 points on a 0 to 100 scale, and improved function by 7 points, with a number needed to treat of 6 for both. The pain reduction held 3 to 6 months after supervised treatment ended. Quality of life did not change (3 small studies) [5].
Population: 10 trials of people with hip osteoarthritis [5].
Sources: [5]
Caveats: No trial could blind participants, and outcomes were self-reported [5].

Physical therapy and exercise therapyModerate

Recovery after knee replacement

Shown to improve function and pain modestly 3 to 4 months after knee replacement, compared with minimal physiotherapy. No long-term benefit was found.

Compared with minimal physiotherapy [13]; home-based physiotherapy [13]

The numbers and sources

Effect: Against minimal physiotherapy, physiotherapy exercise improved function (SMD -0.37, 95% CI -0.62 to -0.12) and pain (SMD -0.45, 95% CI -0.85 to -0.06) at 3 to 4 months. Outpatient and home-based programs did not differ for function or pain, and adding balance or cycling made no difference [13].
Population: 1,739 patients with recent total knee replacement in 18 trials [13].
Sources: [13]
Caveats: Studies were small and underpowered individually and combined [13].

PRP injectionsModerate

Knee osteoarthritis, compared with hyaluronic acid

Shown in pooled trials to give better knee pain, function and WOMAC scores after injection than hyaluronic acid injections.

Compared with hyaluronic acid injections [5]

The numbers and sources

Effect: Across 27 randomized trials, scores after injection were better with PRP than with hyaluronic acid for WOMAC (P < .001), pain on the visual analog scale (P < .01) and the Subjective IKDC knee score (P < .001) [5].
Population: 1,042 patients given PRP and 1,128 given hyaluronic acid, mean age 57.7 and 59.0, mean follow-up about 14 months [5].
Sources: [5]
Caveats: Hyaluronic acid is an active injection, not a placebo. The abstract gives P values without effect sizes [5].

PRP injectionsModerate

Rotator cuff repair, added during surgery

Shown to lower repair failure after rotator cuff surgery, 15.9% against 29.0%, with function gains below the clinically important threshold.

Compared with rotator cuff repair without PRP [12]

The numbers and sources

Effect: Leukocyte-poor PRP injected during arthroscopic repair lowered repair failure to 34 of 213 (15.9%) from 63 of 217 (29.0%) (RR 0.55, 95% CI 0.38 to 0.79). Constant, UCLA and other function scores were higher at 6 to 24 months, by amounts below the minimal clinically important difference. Complications were 9.4% and 7.7% (P = .508) [12].
Population: 470 patients in 5 randomized trials [12].
Sources: [12]
Caveats: Applies to PRP given during surgery, not to injections for tendon pain without surgery.

Red light therapyModerate

Neck pain

Shown to reduce chronic neck pain by about 20 mm on a 100 mm scale compared with placebo, lasting up to 22 weeks after treatment, in one meta-analysis.

The numbers and sources

Effect: Across 16 RCTs (820 patients), pain intensity fell 19.86 mm on a 100 mm visual analogue scale (95% CI 10.04 to 29.68) in 11 trials. In 7 trials with follow-up of 1 to 22 weeks, the reduction was 22.07 mm (95% CI 17.42 to 26.72). In chronic neck pain, the relative risk of improvement was 4.05 (95% CI 2.74 to 5.98) [8].
Population: 820 patients with acute or chronic neck pain in 16 trials [8].
Sources: [8]
Caveats: One meta-analysis, published in 2009, of clinic laser treatment applied to painful sites.

Red light therapyModerate

Tendon pain and plantar fasciitis

Shown to reduce pain in lower-limb tendinopathy and plantar fasciitis by about 13 mm on a 100 mm scale in one meta-analysis, with wide intervals.

The numbers and sources

Effect: Across randomized trials, low-level laser reduced pain by 13.15 mm on a 100 mm scale at the end of therapy (95% CI 7.82 to 18.48) and 12.56 mm at 4 to 12 weeks (95% CI 5.69 to 19.42). Disability improved (SMD 0.39, 95% CI 0.09 to 0.7). Added to exercise at the doses advised by the World Association for Laser Therapy, it reduced pain 18.15 mm more than exercise alone [9].
Population: People with lower extremity tendinopathy or plantar fasciitis in 18 trials, 10 of them against placebo [9].
Sources: [9]
Caveats: The authors note wide confidence intervals, no large trials and no long-term data [9].

Sex hormone binding globulin (SHBG)Moderate

Fractures in older adults

Linked to more fractures at higher levels in older adults, 18% more for each 1 µg/dL higher SHBG across 16 pooled studies.

The numbers and sources

Effect: In a meta-analysis of 16 observational studies, each 1 µg/dL higher SHBG had a relative risk of fracture of 1.18 (95% CI 1.11 to 1.26), with 1.22 (1.12 to 1.33) in men and 1.15 (1.05 to 1.26) in women. By site it was 1.43 (1.23 to 1.65) for hip fracture and 1.31 (1.12 to 1.54) for vertebral fracture [7]. In women aged 65 or older, SHBG of 1.0 µg/dL (34.7 nmol/L) or higher had a relative risk of 2.0 for hip fracture (1.1 to 3.9) and 2.3 for vertebral fracture (1.2 to 4.4) [8]. In older Swedish men, each standard deviation higher SHBG had a hazard ratio for fracture of 1.41 (1.22 to 1.63) [9].
Population: 16 observational studies of older adults [7]. 133 women with later hip fracture and 138 with later vertebral fracture, compared with randomly selected women from the same cohort [8]. 2,639 men with a mean age of 75 [9].
Sources: [7], [8], [9]
Caveats: Observational. One source prints 1.0 µg/dL as 34.7 nmol/L [8]. SHBG rises with age in men [2], and the meta-analysis authors call for study of the mechanisms behind the link [7].

SGLT2 inhibitorsModerate

Gout

Shown to lower gout events by about 45% against placebo in adults with type 2 diabetes or heart failure, in one meta-analysis of five trials.

The numbers and sources

Effect: Across five randomized trials with 1,052 gout-related events, SGLT2 inhibitors were associated with fewer gout outcomes than placebo (HR 0.55, 95% CI 0.45 to 0.67). The effect was smaller with empagliflozin (HR 0.68, 95% CI 0.57 to 0.81) than in the other trials (HR 0.46, 95% CI 0.39 to 0.55), and was not explained by how far uric acid fell [18].
Population: 29,776 patients in five trials, 23,780 with type 2 diabetes [18].
Sources: [18]
Caveats: Not an approved use. Gout was not a primary outcome of the trials, and heterogeneity was moderate (I² 61%) [18].

SNRIsModerate

Chronic back pain and osteoarthritis pain

Shown to lower back pain by about 5 points on a 100-point scale, which reviewers called not clinically important, and osteoarthritis pain by about 10 points.

The numbers and sources

Effect: On a scale from 0 to 100, SNRIs lowered back pain by 5.30 points more than placebo at 3 to 13 weeks (95% CI 3.30 to 7.31) and osteoarthritis pain by 9.72 points (95% CI 6.69 to 12.75). Back pain disability fell by 3.55 points. The authors judged the back pain effect small and not clinically important, and could not exclude a clinically important effect in osteoarthritis [17]. Across chronic pain conditions, duloxetine 60 mg lowered pain intensity against placebo (SMD -0.31, 95% CI -0.39 to -0.24) [15].
Population: 5,318 participants with back pain, sciatica or osteoarthritis in 33 trials [17]; 4,959 participants in 18 studies for pain intensity [15].
Sources: [17], [15], [18]
Caveats: Certainty was moderate for back pain and low for osteoarthritis [17]. Duloxetine is the only SNRI with a chronic musculoskeletal pain approval [18]. Most trials were short and excluded people with low mood [15].

Strength trainingModerateStudied in women

Bone density

Shown to raise hip bone density by about 1% with lower-body strength training in postmenopausal women in a Cochrane review, without a measured drop in fractures.

Compared with control groups in randomized trials [14]; a home-based low-intensity exercise program [15]

The numbers and sources

Effect: Progressive resistance training for the legs raised femoral neck bone density by 1.03% (95% CI 0.24 to 1.82) against controls. Combination programs raised spine density by 3.22%. Fractures did not differ (OR 0.61, 95% CI 0.23 to 1.64) [14]. In women with low bone mass, 8 months of supervised high-intensity resistance and impact training changed lumbar spine density by 2.9% against -1.2%, and femoral neck density by 0.3% against -1.9% [15].
Population: 4,320 postmenopausal women in 43 trials [14]. 101 postmenopausal women with low bone mass, aged about 65 [15].
Sources: [14], [15]
Caveats: Reporting quality in the trials was low [14]. The fracture analysis was small, and the high-intensity trial was one small study with close supervision [15].

Tai chi and qigongModerate

Knee osteoarthritis, against no program or attention control

Shown to ease knee osteoarthritis pain and stiffness and improve function in one meta-analysis of 16 trials. A small trial found more relief than wellness education and stretching.

Compared with control groups the source abstract does not describe [8]; wellness education and stretching, as an attention control [9]

The numbers and sources

Effect: Tai chi improved pain (SMD -0.69, 95% CI -0.95 to -0.44), stiffness (SMD -0.59), physical function (SMD -0.92, 95% CI -1.16 to -0.69), dynamic balance (SMD 0.69) and physical and mental quality of life [8]. In 40 adults, WOMAC pain improved more with tai chi than with the attention control at 12 weeks (mean difference -118.80 mm, 95% CI -183.66 to -53.94), along with function, chair stand time and depression scores [9].
Population: 986 patients with knee osteoarthritis in 16 trials [8]. 40 adults with symptomatic knee osteoarthritis, mean age 65 [9].
Sources: [8], [9]
Caveats: Strength of evidence for the main outcomes was low or moderate [8]. Patients knew their group [9].

Testosterone therapyModerateStudied in men

Bone density in men

Shown to raise spine bone density about 3.7% in middle-aged and older men, with no change at the hip. Density is not fracture.

The numbers and sources

Effect: Lumbar spine bone mineral density rose 3.7% (95% CI 1.0 to 6.4) over placebo. Femoral neck density did not differ. Bone resorption markers fell by an SMD of -0.6 (95% CI -1.0 to -0.2) [2].
Population: Middle-aged and older men, 29 RCTs [2].
Sources: [2]
Caveats: Density is not fracture. The one large trial that counted fractures found more of them on testosterone (see Risks) [4].

Thiazide diureticsModerate

Fractures

Linked to about 14% fewer fractures in cohort studies. In one randomized trial, chlorthalidone brought fewer hip and pelvic fractures than amlodipine or lisinopril.

The numbers and sources

Effect: In 11 cohort studies, thiazide users had fewer fractures of any kind than non-users (RR 0.86, 95% CI 0.80 to 0.93) and fewer hip fractures (RR 0.82). The effect was weaker in the higher-quality studies [15]. In a secondary analysis of ALLHAT, people randomized to chlorthalidone had fewer hip and pelvic fractures during the trial than those on amlodipine or lisinopril (HR 0.79, 95% CI 0.63 to 0.98). Over the trial plus later follow-up the difference was not significant (HR 0.87, 95% CI 0.74 to 1.03) [16].
Population: 2,193,160 participants in 11 cohort studies [15]; 22,180 ALLHAT participants with a mean age of 70.4, followed up to 8 years in the trial [16].
Sources: [15], [16]
Caveats: The cohort studies are observational. The randomized evidence is a secondary analysis against other drugs, not placebo [16].

Thyroid stimulating hormone (TSH)Moderate

Fractures

Linked to more hip and other fractures at TSH below 0.45 mIU/L, most at TSH below 0.10, with no link at raised TSH.

The numbers and sources

Effect: Against a TSH of 0.45 to 4.49 mIU/L, people with TSH below 0.45 mIU/L and normal thyroxine had a hazard ratio of 1.36 for hip fracture (95% CI 1.13 to 1.64) and 1.28 for any fracture (1.06 to 1.53). At TSH below 0.10 mIU/L the hazard ratios were 1.61 for hip fracture (1.21 to 2.15), 1.98 for any fracture (1.41 to 2.78) and 3.57 for spine fracture (1.88 to 6.78). No association was found between TSH of 4.50 to 19.99 mIU/L and fracture [20].
Population: 70,298 adults in 13 prospective cohorts, 762,401 person-years, 2,975 hip fractures [20].
Sources: [20]
Caveats: Observational. The authors state that trials are lacking on whether treating a low TSH prevents fractures [20].

Vitamin A (retinol) levelModerate

Fracture

Linked to more hip fractures at both high and low blood retinol in a meta-analysis of four cohorts, with no link to fractures overall.

The numbers and sources

Effect: In a meta-analysis of prospective studies, both high and low blood retinol went with more hip fractures (RR 1.87, 95% CI 1.31 to 2.65, and RR 1.56, 1.09 to 2.22), in a U-shaped relation. The risk of all fractures taken together did not differ by blood retinol [10]. In one of the cohorts, Swedish men in the highest fifth of serum retinol (above 75.62 µg/dL) had a rate ratio of 1.64 (1.12 to 2.41) for any fracture and 2.47 (1.15 to 5.28) for hip fracture against the middle fifth (62.16 to 67.60 µg/dL). Men above the 99th percentile (103.12 µg/dL) had about seven times the fracture risk of men with lower levels [4].
Population: 8,725 participants in four prospective studies of blood retinol [10]. 2,322 men aged 49 to 51 at enrollment, followed for 30 years, with 266 fractures [4].
Sources: [10], [4]
Caveats: Observational, from four cohorts. The Swedish cohort measured retinol once, at about age 50, in men only [4].

Whole-body cryotherapyModerate

Ankylosing spondylitis

Shown to improve pain, function and disease activity scores in people with ankylosing spondylitis in a meta-analysis of 5 studies. C-reactive protein did not change.

The numbers and sources

Effect: Across 5 studies (310 patients), cryotherapy improved the BASDAI disease activity index (p < 0.001), ASDAS (p = 0.015), BASFI function index (p = 0.006) and pain scores (p = 0.005). C-reactive protein did not change (p = 0.684) [11].
Population: 310 people with ankylosing spondylitis in 5 studies with varied temperatures and session lengths of 80 seconds to 3 minutes [11].
Sources: [11]
Caveats: Five studies, with protocols that varied. The abstract gives p values without effect sizes, and the authors call for studies of tolerance and long-term results.

YogaModerate

Chronic low back pain, against no exercise

Shown to improve back function and pain slightly against no exercise in a Cochrane review, by amounts below the thresholds the reviewers set as clinically important.

Compared with waiting list, usual care or education [1]; a self-care book [2]; an educational book and newsletters [3]

The numbers and sources

Effect: At 3 months, back function improved 1.69 points on the 24-point Roland-Morris scale (95% CI -2.73 to -0.65, 11 trials, 1,155 participants, minimal clinically important difference 5 points) and pain fell 4.53 points on a 0 to 100 scale (95% CI -6.61 to -2.46, 9 trials, clinically important difference 15 points). More people rated their back pain improved or resolved (risk ratio 2.33, 95% CI 1.46 to 3.71) [1]. In 228 adults, function at 12 weeks was 2.5 points better with 12 yoga classes than with a self-care book (95% CI -3.7 to -1.3), and still 1.8 points better at 26 weeks [2]. In 320 mostly low-income adults, yoga was not superior to an educational book for function or pain at 12 weeks, and 21 percentage points fewer yoga participants used pain medication [3].
Population: 2,223 adults in 21 trials, mostly women in their 40s or 50s [1]. 228 primary care patients [2]. 320 racially diverse adults at a safety-net hospital and 7 health centers [3].
Sources: [1], [2], [3]
Caveats: Every trial was at high risk of performance and detection bias, and no trial compared yoga with sham yoga. The reviewers judged it unlikely that blinded comparisons would find a clinically important benefit [1].

YogaModerate

Flexibility and leg strength in older adults

Shown to improve lower body flexibility and leg strength in older adults against both inactive controls and active exercise controls, in one meta-analysis.

Compared with inactive controls such as a waiting list or education booklets [7]; active controls such as walking or chair aerobics [7]

The numbers and sources

Effect: Against inactive controls, lower body flexibility improved (effect size 0.5) and lower limb strength improved (0.45). Against active controls, lower limb strength improved (0.49) and lower body flexibility improved (0.28) [7].
Population: Older adults with a mean age of 60 or above, not recruited for any condition, in 17 trials of physical function [7].
Sources: [7]
Caveats: The abstract gives no confidence intervals.

BisphosphonatesLimited

New fractures and survival after a hip fracture

One trial found yearly zoledronic acid after hip fracture surgery lowered new fractures, 8.6% against 13.9%, and deaths, 9.6% against 13.3%, over about 2 years.

Compared with placebo, with calcium and vitamin D in both groups [7]

The numbers and sources

Effect: New clinical fractures occurred in 8.6% on zoledronic acid and 13.9% on placebo, a 35% reduction. New clinical spine fractures were 1.7% against 3.8%, and non-vertebral fractures 7.6% against 10.7%. Deaths from any cause were 9.6% against 13.3%, 28% fewer [7].
Population: 2,127 adults, mean age 74.5, given the first infusion within 90 days of hip fracture repair, median follow-up 1.9 years [7].
Sources: [7]
Caveats: One trial. The pooled analysis of mortality below did not confirm a drop in deaths across trials [8].
Why limited: This rests on one trial, of 2,127 people. Strength counts independent studies that agree, so one trial is limited at any size.

BoronLimited

Joint pain and arthritis

Small trials found boron supplements eased knee discomfort over 2 weeks and lowered inflammation markers in arthritis. Several were short, and one was run by the maker.

The numbers and sources

Effect: In 60 adults with self-reported knee discomfort, calcium fructoborate 110 mg twice a day lowered McGill Pain Questionnaire scores by 8.9 points and WOMAC scores by 13.73 points more than placebo at day 14 (both p < 0.0001), with no change on placebo [10]. In 60 adults with knee osteoarthritis, 15 days of calcium fructoborate improved C-reactive protein, fibrinogen and sedimentation rate against placebo, with no serious adverse events [11]. In 72 people with rheumatoid arthritis on etanercept, calcium fructoborate or borax (6 mg of boron a day) for 60 days improved disease activity scores and lowered ESR, hsCRP, IL-1α, IL-6 and TNF-α against placebo [12]. A 1994 review reported a 20-person trial in osteoarthritis in which 50% improved on 6 mg of boron a day against 10% on placebo [13].
Population: 60 adults aged 35 to 65 with knee discomfort in Mexico [10]. 60 middle-aged adults with knee osteoarthritis [11]. 72 adults with rheumatoid arthritis in Iraq [12]. 20 adults with osteoarthritis [13].
Sources: [10], [11], [12], [13]
Caveats: Trials ran 2 weeks to 2 months. The knee discomfort trial was designed by FutureCeuticals, which supplied the calcium fructoborate and employed most authors [10]. The 20-person trial is known only from its author's review [13]. No meta-analysis was located.

BPC-157Grey marketLimited

Knee pain

One small uncontrolled study found 14 of 16 patients reported knee pain relief after BPC-157 injections, some combined with thymosin beta-4.

Compared with the same patients' recall of their pain before the injection, with no other group [14]

The numbers and sources

Effect: A chart review at one US clinic reached 16 patients by phone 6 months to 1 year after a knee injection. Of 12 who received BPC-157 alone, 11 reported significant improvement in knee pain. Of 4 who received BPC-157 with thymosin beta-4, 3 reported significant improvement. Overall, 14 of 16 (87.5%) reported relief. No tools were used to measure function, stiffness, quality of life or daily activities [14]. A 2026 systematic review judged the study of poor methodological quality and rated the certainty of evidence very low [9].
Population: 16 patients with knee pain of several causes, from a chart review covering 2019 to 2020 [14].
Sources: [14], [9], [10]
Caveats: Retrospective, uncontrolled and unblinded, with pain recalled in a phone survey [14]. With no comparison group, the study cannot separate an effect of the injection from placebo response or natural recovery. A 2026 review describes human musculoskeletal evidence for BPC-157 as limited to this one small case series without validated outcome measures [10].

Chiropractic and spinal manipulationLimited

Cervicogenic headache

Limited evidence: two low-quality trials suggested several sessions of neck manipulation eased cervicogenic headache pain more than massage.

Compared with massage [7]

The numbers and sources

Effect: For chronic cervicogenic headache, multiple sessions of neck manipulation may improve pain and function more than massage at short and intermediate-term follow-up [7].
Population: 125 participants in 2 trials [7].
Sources: [7]
Caveats: Low-quality evidence from two trials, with no sham comparison [7].

CollagenLimited

Collagen synthesis and tendon or ligament injury

Limited evidence: 15 g of vitamin C-enriched gelatin doubled a blood marker of collagen synthesis in 8 men. Injury studies were sparse and low certainty.

The numbers and sources

Effect: In 8 healthy men, 15 g of vitamin C-enriched gelatin taken 1 hour before 6 minutes of rope skipping doubled a blood marker of collagen synthesis (amino-terminal propeptide of collagen I). The abstract reports this doubling for the 15 g dose, not the 5 g dose [11]. A systematic review of 15 RCTs reported collagen synthesis rates were raised with 15 g a day [3]. A 2026 review of 6 studies in tendon and ligament tears, repairs and related conditions found collagen-based treatments associated with better pain, function, analgesic use or imaging results [12].
Population: 8 healthy men, crossover [11]; 15 RCTs, mostly recreational athletes [3]; people with tendon or ligament tears, after ACL reconstruction or rotator cuff repair, and with tendinopathy or ankle instability [12].
Sources: [3], [11], [12]
Caveats: Blood markers and engineered ligaments in a lab, not injury rates [11]. In the injury review, most oral products contained several ingredients besides collagen, and its authors rated the evidence sparse and low certainty [12].

Daily stepsLimited

Falls

Limited evidence: one meta-analysis linked 7,000 steps a day to 28% fewer falls than 2,000, at very low certainty.

Compared with 2,000 steps a day [4]

The numbers and sources

Effect: Against 2,000 steps a day, 7,000 steps was linked to a 28% lower risk of falls (HR 0.72, 95% CI 0.65 to 0.81), with an inflection at about 5,000 to 7,000 steps [4].
Population: 4 cohort studies [4].
Sources: [4]
Caveats: The review rated certainty very low [4]. People with poor balance or past falls walk less, which is reverse causation.

Dry needlingLimited

Knee osteoarthritis, other comparisons

Limited evidence: pooled small trials found less osteoarthritis pain and better function in the short term, with no difference at medium or long term.

Compared with non-drug treatments, sham techniques or no added treatment, pooled together [13]; control, sham or placebo treatment [14]

The numbers and sources

Effect: Across 7 trials, dry needling lowered pain (SMD -0.76, 95% CI -1.24 to -0.29, I² 74%) and improved physical function (SMD -0.98, 95% CI -1.54 to -0.42, I² 75%) in the short term, with no difference at medium or long term [13]. A review of 9 trials found contradictory results for trigger point needling. Periosteal stimulation improved pain and function right after treatment, and intramuscular electrical stimulation lowered pain with marked heterogeneity [14].
Population: 291 people with hip or knee osteoarthritis in 7 trials [13]. 9 trials of knee osteoarthritis [14].
Sources: [13], [14]
Caveats: Evidence was rated very low quality for risk of bias, heterogeneity and imprecision [13]. Sham and no-treatment comparisons were pooled together.

Ezetimibe and fenofibrateLimited

Fenofibrate: uric acid and gout

One trial's later analysis found fenofibrate lowered uric acid about 20% and nearly halved first gout attacks over 5 years in type 2 diabetes, 2% against 3%.

Compared with placebo [15]

The numbers and sources

Effect: Uric acid fell 20.2% during the fenofibrate run-in and stayed 20.1% lower than placebo at 1 year. First gout events were 81 (2%) on fenofibrate and 151 (3%) on placebo over 5 years (HR 0.54, 95% CI 0.41 to 0.70), and all gout events halved (HR 0.48) [15].
Population: 9,795 adults aged 50 to 75 with type 2 diabetes [15].
Sources: [15]
Caveats: A post hoc analysis of a heart trial, not a gout trial.
Why limited: This rests on one trial, of 9,795 people. Strength counts independent studies that agree, so one trial is limited at any size.

Follicle-stimulating hormone (FSH)LimitedStudied in women

Bone loss across the menopausal transition

One cohort found FSH levels and their change over 4 years predicted spine and hip bone loss through the menopausal transition, while estradiol changes did not.

The numbers and sources

Effect: Over 4 years, lumbar spine bone density fell 5.6% in women who reached natural postmenopause and 3.2% in late perimenopause. Baseline FSH, later FSH and their interaction predicted the loss. With a baseline FSH below 25 mIU/mL, a follow-up FSH above 70 predicted a 4-year spine loss of 0.05 g/cm2. Estradiol below 35 pg/mL went with lower bone density, but yearly estradiol changes did not predict loss [12].
Population: 2,311 premenopausal or early perimenopausal women at five sites of the Study of Women's Health Across the Nation, examined yearly for five visits [12].
Sources: [12]
Caveats: Observational, from one cohort. FSH rises as ovarian function declines, so it may mark the stage of the transition and not act on bone itself.

GroundingLimited

Pain

Limited evidence: two small studies reported less pain while grounded, one in 16 massage therapists compared with ungrounded weeks and one uncontrolled study of 12 adults.

Compared with weeks without grounding, in the same people [3]; before grounding, in the same people [2]

The numbers and sources

Effect: In a stepped wedge design, massage therapists reported significant decreases in pain, fatigue, tiredness and depressed mood and increases in physical function and energy while grounded at work and during sleep, compared with weeks not grounded [3]. Self-reported pain was reduced or eliminated in nearly all of 12 adults who slept grounded for 8 weeks [2].
Population: 16 massage therapists, mean age 42.8, at one wellness center in California [3]. 12 adults [2].
Sources: [3], [2]
Caveats: Sixteen and 12 people. The massage therapist trial was led by an author who, in another paper, disclosed work as a contractor for the company that sponsors grounding research [3][4]. Outcomes were self-reported.

Heat therapyLimited

Acute low back pain, compared with ibuprofen and acetaminophen

One trial found a heat wrap eased acute low back pain more than ibuprofen 1,200 mg or acetaminophen 4,000 mg a day, on day 1 and days 3 to 4.

Compared with ibuprofen and acetaminophen [8]

The numbers and sources

Effect: Day 1 pain relief averaged 2 with the heat wrap, 1.51 with ibuprofen and 1.32 with acetaminophen. Relief on days 3 to 4 averaged 2.61, 1.68 and 1.95. Trunk flexibility and disability also improved more with heat [8].
Population: 371 adults with acute nonspecific low back pain, 113 on heat, 106 on ibuprofen and 113 on acetaminophen [8].
Sources: [8]
Caveats: Only the investigator was blinded, and the placebo and unheated-wrap groups (20 and 19 people) were there for blinding, not for comparison [8].

Heat therapyLimited

Acute low back pain, heat with exercise

One trial found a heat wrap plus exercise restored function in 72% of people with acute low back pain, against about 20% with heat, exercise or a booklet alone.

Compared with an education booklet [9]; exercise alone [9]; a heat wrap alone [9]

The numbers and sources

Effect: Two days after 5 days of treatment, 72% of the heat plus exercise group had returned to pre-injury function, against 20% with heat alone, 20% with exercise alone and 19% with the booklet. Function improved 175% more than with the booklet [9]. A Cochrane review reports the same finding of added benefit from exercise [7].
Population: 100 adults, mean age 31, with low back pain of less than 3 months [9].
Sources: [9], [7]
Caveats: One small trial with about 25 people per group [9].

HomeopathyLimited

Fibromyalgia

Limited evidence: four small trials found lower pain and fewer tender points than placebo, results the review authors call preliminary.

Compared with placebo [19]

The numbers and sources

Effect: Against placebo, homeopathy lowered tender point count (SMD -0.42, 95% CI -0.78 to -0.05), pain intensity (SMD -0.54, 95% CI -0.97 to -0.10) and fatigue (SMD -0.47, 95% CI -0.90 to -0.05) [19].
Population: 4 randomized trials, plus case reports and observational studies [19].
Sources: [19]
Caveats: Few small trials. The authors write that any conclusions are preliminary [19].

L-arginineLimitedStudied in children

Height growth in children with short stature

Limited evidence: oral arginine raised growth speed about 0.6 cm a year in children with idiopathic short stature, in low-quality studies.

The numbers and sources

Effect: Oral arginine for 6 or 12 months raised growth velocity (SMD 0.66, 95% CI 0.21 to 1.12), equal to 0.59 cm a year (95% CI 0.19 to 1.00). The effect was not significant in studies giving it for 12 months only [14].
Population: Prepubertal children with idiopathic short stature, in clinical trials and cohort studies [14].
Sources: [14]
Caveats: The authors found only low-quality studies [14]. Growth hormone levels were not reported in the abstract.

Massage therapyLimited

Knee osteoarthritis

One trial of 222 adults found weekly massage improved knee osteoarthritis scores more than light touch or usual care at 8 weeks, with no difference at 52 weeks.

Compared with light touch [7]; usual care [7]

The numbers and sources

Effect: At 8 weeks, WOMAC global scores improved by 8.16 points more than light touch (95% CI -13.50 to -2.81) and 9.55 points more than usual care (95% CI -14.66 to -4.45), with gains in pain, stiffness and function. At 52 weeks, change did not differ across groups (P = 0.707) [7].
Population: 222 adults with knee osteoarthritis [7].
Sources: [7]
Caveats: One trial. Light touch was the closest thing to a sham, and participants knew which hands-on treatment they had.

Muscle relaxantsLimited

Methocarbamol for acute low back pain

One trial found 44% on methocarbamol stopped treatment early with complete pain relief, against 18% on placebo, in acute low back pain with spasm.

The numbers and sources

Effect: Over up to 8 days, 44% on methocarbamol and 18% on placebo ended treatment early because pain was gone, and 19% against 52% stopped because treatment was judged not to be helping. At the final visit 67% on methocarbamol and 35% on placebo rated the treatment as having helped. Adverse events were reported by 5 patients [7].
Population: 202 adults with acute low back pain for at least 24 hours with spasm and restricted movement, 98 on methocarbamol and 104 on placebo [7].
Sources: [7]
Caveats: One trial, run in 2002 and analyzed after the fact for publication in 2015 [7]. A systematic review found very limited or inconsistent data for methocarbamol against placebo [4].

Ozone therapyLimited

Low back pain and herniated disc, compared with sham injections

Two small double-blind trials found ozone injections beside the spine eased back pain and disability more than injections of oxygen or very dilute ozone.

Compared with paraspinal injections of pure oxygen [2]; paravertebral injections of very dilute ozone, 0.1 µg/mL [3]

The numbers and sources

Effect: In chronic discogenic low back pain, pain on the visual analog scale fell 5.6 points with ozone and 2.4 with pure oxygen, and the Oswestry disability score fell 19.07 and 7.8 points (P < 0.001), with exercise in both groups [2]. In acute lumbar disc herniation, pain and disability scores were lower with ozone than with placebo at 15 and 30 days and one month after treatment ended (P < 0.05), added to conservative treatment [3].
Population: 30 adults with MRI-confirmed chronic discogenic low back pain [2]. 38 adults with acute lumbar disc herniation [3].
Sources: [2], [3]
Caveats: Two small single-center trials with follow-up of 3 months or less. Pure oxygen and dilute ozone are injected gases themselves, so neither control is inert.

Ozone therapyLimited

Knee osteoarthritis, compared with placebo

Limited evidence: one double-blind trial and three reviews found less knee osteoarthritis pain than placebo, and an umbrella review rated every review critically low in confidence.

Compared with placebo injections [6][7]

The numbers and sources

Effect: After 8 weeks of injections, pain was 2.16 points lower than placebo on the visual analog scale (95% CI 0.42 to 3.89), and the Lequesne index, WOMAC scores and several quality of life scores improved more. The Timed Up and Go test did not differ [6]. An umbrella review of 8 reviews covering 15 trials found 3 reporting significant pain reduction against placebo, while 5 found ozone no better than other therapies, with less consistent effects on joint function [7].
Population: 98 adults with symptomatic knee osteoarthritis in Brazil [6]. 8 systematic reviews of 15 trials with 3,685 patients [7].
Sources: [6], [7]
Caveats: All 8 reviews had critically low confidence ratings on AMSTAR 2 [7]. The placebo trial lasted 8 weeks [6].

Physical therapy and exercise therapyLimited

Knee osteoarthritis, compared with a steroid injection

One trial of 156 adults found less knee pain and disability at 1 year with physical therapy than with a glucocorticoid injection.

Compared with intra-articular glucocorticoid injection [8]

The numbers and sources

Effect: At 1 year, WOMAC scores (0 to 240) were 37.0 with physical therapy and 55.8 with injection, a difference of 18.8 points (95% CI 5.0 to 32.6) [8].
Population: 156 patients with knee osteoarthritis, mean age 56, in the US Military Health System [8].
Sources: [8]
Caveats: One trial, without blinding of participants.

RetatrutideGrey marketLimited

Knee osteoarthritis pain

One trial found knee pain scores fell about 1.5 points more than placebo in people with knee osteoarthritis. The placebo group also improved.

The numbers and sources

Effect: In people with knee osteoarthritis, the WOMAC pain score changed by -3.4, -3.9 and -4.1 on 4, 9 and 12 mg, against -2.5 on placebo by intention to treat, differences of -1.4 and -1.6 points for 9 mg and 12 mg [1].
Population: 574 participants with knee osteoarthritis within TRIUMPH-1 [1].
Sources: [1]
Caveats: A subgroup of one trial. The placebo group also improved by 2.5 points.

SaunaLimited

Low back pain

Limited evidence: pain scores fell from a median of 5 to 3 in 37 people after 10 sauna sessions, in a study with no control group.

The numbers and sources

Effect: After 10 sessions of 15 minutes at 90°C over 5 days, median pain on a verbal numerical rating scale fell from 5 (range 2 to 8) to 3 (range 0 to 8), and median Oswestry disability scores fell from 12 to 8 (P < 0.001 for both) [21].
Population: 37 consecutive patients over 20 years of age with low back pain [21].
Sources: [21]
Caveats: No control group, so the change cannot be separated from natural recovery or expectation. One week of follow-up.

Sitting timeLimited

Neck and shoulder pain

One trial in office workers found less neck and shoulder pain after 3 months of a program to cut sitting, and no change in back pain.

Compared with control offices whose workers behaved as usual [12]

The numbers and sources

Effect: At 3 months, the odds of neck and shoulder pain were lower in the program group (OR 0.52, 95% CI 0.30 to 0.92), with no difference at 1 month. Back and limb pain did not differ. Total pain score fell slightly [12].
Population: 317 office workers in 19 offices [12].
Sources: [12]
Caveats: Pain was a secondary outcome, self-reported by workers who knew their group.

Tai chi and qigongLimited

Falls, against other exercise

One trial of 670 adults at high risk of falling found a tai chi balance program cut falls 58% against stretching and 31% against a multimodal exercise program.

Compared with stretching exercises [4]; multimodal exercise combining balance, aerobic, strength and flexibility work [4]

The numbers and sources

Effect: Over 6 months, there were 152 falls in 85 people with tai chi, 218 falls in 112 people with multimodal exercise and 363 falls in 127 people with stretching. The incidence rate ratio was 0.42 (95% CI 0.31 to 0.56) for tai chi against stretching and 0.69 (95% CI 0.52 to 0.94) against multimodal exercise [4].
Population: 670 community-dwelling adults aged 70 and over who had fallen in the past year or had impaired mobility, mean age 77.7, 65% women, in Oregon [4].
Sources: [4]
Caveats: One single-blind trial of a program tailored for balance, not classic tai chi forms [4].

Total testosteroneLimitedStudied in men

Fractures in a trial that raised testosterone

One trial found more fractures, not fewer, when testosterone was raised in men with low levels: 3.5% against 2.5% on placebo over 3 years.

The numbers and sources

Effect: In a fracture substudy of the same cardiovascular trial, a clinical fracture occurred in 91 of 2,601 men on testosterone gel (3.50%) and 64 of 2,603 on placebo (2.46%), a hazard ratio of 1.43 (95% CI 1.04 to 1.97). Fracture incidence also appeared higher with testosterone for all other fracture end points [24].
Population: 5,204 men aged 45 to 80 with two morning testosterone levels below 300 ng/dL, median follow-up 3.19 years [24].
Sources: [24]
Caveats: A marker that tracks an outcome is not always one whose correction changes it. The authors note that testosterone treatment had improved bone density in earlier studies, and this result ran the other way [24].
Why limited: This rests on one trial, of 5,204 people. Strength counts independent studies that agree, so one trial is limited at any size.

Whole-body cryotherapyLimited

Frozen shoulder

One trial found greater gains in pain and shoulder movement when cryotherapy was added to physical therapy in 30 people with adhesive capsulitis.

The numbers and sources

Effect: Pain, active range of motion in four directions and a shoulder function score improved more in the group that added whole-body cryotherapy to physical therapy and joint mobilization (P < .01) [16].
Population: 30 patients with adhesive capsulitis of the shoulder, randomized to two groups [16].
Sources: [16]
Caveats: Thirty people, no sham, and the abstract gives no sizes.

Whole-body cryotherapyLimited

Chronic low back pain

Limited evidence: pain and disability scores fell after a course of sessions in two studies with no control group.

The numbers and sources

Effect: In 41 adults with chronic low back pain, pain and disability scores fell significantly over 20 sessions in 5 weeks (p < 0.001) [17]. In 30 patients with lumbar spine osteoarthritis who also did therapeutic exercise, back pain fell significantly (p < 0.0001) and plasma cortisol rose (p = 0.0009) [18].
Population: 41 adults aged 18 to 65 [17]. 30 patients with lumbar spine osteoarthritis [18].
Sources: [17], [18]
Caveats: Neither study had a control group, and one combined cryotherapy with exercise [18].

Reported risks71

GnRH drugs for fibroidsStrongVery common

Bone density loss

Shown to lower bone density, more over time. With add-back, spine density fell 0.8% to 1.5% at 12 months, and stayed below baseline 12 months after stopping elagolix.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: On Myfembree, lumbar spine density changed -0.23% against +0.18% on placebo at 6 months, and -0.80% at 12 months. A drop of more than 3% at 12 months occurred in 23% on Myfembree and in 17.4% of untreated women with fibroids in an observational cohort. Four women on Myfembree had low-trauma fractures [2]. On Oriahnn, spine density fell 0.7% against 0.1% on placebo at 6 months, and 1.5% at 12 months, with a drop of more than 3% in 27%. Twelve months after stopping, spine density was still 0.72% below baseline, and loss continued in 24%, partly recovered in 46% and fully recovered in 30%. Fractures occurred in 7 of 453 women on Oriahnn (1.5%) and 1 of 196 on placebo (0.5%) [6]. Without add-back, spine density fell 4.8% at 12 months on elagolix [9], 2.7% after 3 months of leuprolide in a fibroid trial, and 6.3% at 52 weeks of leuprolide in endometriosis trials [10]. Pooled trials found more bone loss on elagolix than on placebo [7][8].
Population: 254 women on Myfembree and 256 on placebo, 163 followed to 12 months [2]; 305 women on Oriahnn and 150 on placebo, 175 at 12 months [6]; women in the elagolix, leuprolide and pooled trials [7][8][9][10].
Sources: [2], [6], [9], [10], [7], [8]
Caveats: Both labels state the effect of these losses on long-term bone health and fracture risk in premenopausal women is unknown, and limit use to 24 months [2][6].

Chiropractic and spinal manipulationModerateVery common

Soreness, stiffness and headache after treatment

Shown in 50% to 67% of patients in large case series, as brief soreness, stiffness or headache. Pooled trials found mild events no more often than with control treatments.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Minor, transient adverse events such as increased pain, muscle stiffness and headache were reported 50% to 67% of the time in large case series [6]. In 14 trials of neck manipulation, the rate of adverse events did not differ from control interventions, and every reported event was mild [15]. In one trial powered to look at harm, adverse events did not differ from sham manipulation (RR 1.24, 95% CI 0.85 to 1.81) [3].
Population: Large case series of patients treated with manipulation [6]. 14 randomized trials [15]. 183 participants in one trial [3].
Sources: [6], [15], [3]
Caveats: Most trials did not record adverse events systematically [3].

IsotretinoinModerateVery common

Joint, back and muscle symptoms

Shown in trials to cause joint or muscle symptoms in about 16%, with back pain in 29% of teenagers and marked creatine kinase rises in 24%.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: About 16% in a clinical trial had musculoskeletal symptoms, including joint pain. In a trial of teenagers, 29% had back pain, severe in 14% of those cases, and 22% had joint pain. Marked creatine kinase rises (350 U/L or more) occurred in about 24% of 924 trial participants, some linked to vigorous exercise, and rare rhabdomyolysis has been reported [3].
Population: 358 patients aged 12 and older in a pediatric trial, and 924 participants in another trial [3].
Sources: [3]
Caveats: Uncontrolled label figures with no placebo rate.

Physical therapy and exercise therapyModerateVery common

Pain flare-ups and soreness

Shown to raise minor adverse events by 19% in pooled trials, about one extra for every 6 people. Most were increased pain or muscle soreness.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Across exercise therapy trials, non-serious adverse events rose (RR 1.19, 95% CI 1.09 to 1.30), with one extra event for every 6 people treated [14]. In the hip osteoarthritis trial, 19 of 46 people (41%) on physical therapy and 7 of 49 (14%) on sham reported mild adverse effects [6]. In back pain trials that measured harms systematically, there was a median of 0.14 adverse effects per participant with exercise and 0.12 with comparison treatments, mostly minor, such as muscle soreness [2]. In knee and hip osteoarthritis trials, reported adverse events were increased knee, hip or low back pain [4][5].
Population: 38,517 participants in 375 trials [14]. 95 adults with hip osteoarthritis [6]. 249 back pain trials [2]. Osteoarthritis trials [4][5].
Sources: [14], [6], [2], [4], [5]
Caveats: The 41% comes from one trial of a multi-part program [6]. Many trials did not report adverse events [2].

PRP injectionsModerateVery common

Pain, swelling and other injection reactions

Shown in knee osteoarthritis trials to cause complications in 18.66% against 9.14% with comparison injections, more often than saline, almost all mild or moderate.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Across 24 knee osteoarthritis trials, complications occurred in 18.66% given PRP and 9.14% given comparison injections, a number needed to harm of 11. Odds were higher than with saline (OR 4.88) and did not differ from hyaluronic acid (OR 1.33) or corticosteroids. One patient reported severe pain, and the rest were mild to moderate and self-limiting [17]. In the Achilles trial, injection site discomfort occurred in 97 of 121 given PRP and 73 of 119 given sham, swelling in 56 and 52, and bruising in 48 and 49 [7]. In tennis elbow trials, adverse events occurred in 19% with PRP or blood and 17% with placebo (RR 1.14, 95% CI 0.76 to 1.72) [6].
Population: 2,751 patients in 24 knee trials [17]. 240 adults [7]. 425 participants in 5 elbow trials [6].
Sources: [17], [7], [6]
Caveats: The higher rate against saline was seen in knee trials. In tennis elbow trials the difference from placebo was not significant [6].

Loop diureticsLimitedVery common

Raised uric acid and gout

Shown to raise uric acid, about 1.2 mg/dL on torsemide 10 mg, with high uric acid in 18.4% of bumetanide patients tested. Studies linked diuretics to gout, inconsistently.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Torsemide 10 mg a day for 6 weeks raised serum uric acid a mean 1.2 mg/dL, and the change reversed after stopping [1]. High uric acid was reported in 18.4% of patients tested on bumetanide [15]. A systematic review found a trend toward more acute gout attacks with loop and thiazide diuretics, with inconsistent size and independence [16].
Population: Hypertensive patients in torsemide trials [1]; bumetanide-treated patients [15]; 2 trials, 6 cohort and 5 case-control studies [16].
Sources: [1], [15], [16]
Caveats: The two trials in [16] tested thiazide diuretics, not loop diuretics.

Progestins for endometriosis and heavy bleedingLimitedStudied in womenVery common

Bone density with dienogest

Mixed results: over 24 weeks, dienogest left spine bone density unchanged in two trials. In 60 women taking it for a year or more after surgery, spine density fell 2.7%.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Spine bone density changed +0.25% on dienogest and -4.04% on leuprolide over 24 weeks [9]. In 140 women measured in a placebo-controlled trial, dienogest had no effect on bone density at 24 weeks [8]. After surgery for endometriomas, spine density fell 2.2% at 6 months and 2.7% at 1 year on dienogest, femoral neck density fell 2.8% at 1 year, and 45 of 60 women (75%) had a significant fall. Density did not fall further in the second year in 24 women [17].
Population: 252 women [9]. 140 women [8]. 60 women aged about 30 after surgery, treated for a mean of 18.6 months [17].
Sources: [9], [8], [17]
Caveats: The longer study had no control group [17].

TopiramateLimitedStudied in childrenVery common

Bone density and growth in children

One small trial in children found lower bone density gains on topiramate, with 21% losing a clinically important amount against none on the comparison drug, and slower weight gain.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: In a one-year active-controlled trial, topiramate lowered lumbar spine and whole-body bone mineral density, and 21% had a clinically important fall (Z score change of -0.5 or more) against 0 in the control group. Serum parathyroid hormone and 25-hydroxyvitamin D fell, and urinary calcium rose. Mean annual weight gain was lower than on the comparison drug, with a similar trend for height [3].
Population: 63 children with epilepsy, as reported in the label [3].
Sources: [3]
Caveats: Too small and short to judge fracture risk [3].

Osteopathy and osteopathic manipulative treatmentModerateCommon

Pain or discomfort after treatment

Shown at 2.5% of office visits, mostly as pain or discomfort, in patients asked right after treatment. Safety outcomes did not differ from sham in the largest back pain trial.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: Patients felt worse or much worse right after treatment at 45 of 1,847 visits, an adverse event rate of 2.5% (95% CI 1.3% to 4.7%). Pain or discomfort was the most common type (0.9%). Women reported adverse events more often than men (OR 13.9) [18]. In the largest back pain trial, safety outcomes did not differ between osteopathic treatment and sham [3]. Most systematic reviews reported no adverse events [17].
Population: 884 patients at 1,847 visits [18]. 455 adults [3]. 9 systematic reviews [17].
Sources: [18], [3], [17]
Caveats: Effects were recorded before patients left the office, so later effects were not captured [18].

SGLT2 inhibitorsModerateCommon

Leg and foot amputation, a former FDA boxed warning on canagliflozin

Mixed results: one canagliflozin trial program found amputations in 3.3% against 1.5% on placebo, while a later canagliflozin trial and pooled trials of other drugs found no clear rise.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: In the CANVAS program, amputation occurred at 6.3 against 3.4 participants per 1,000 patient-years (HR 1.97, 95% CI 1.41 to 2.75), mostly at the toe or metatarsal [13]. On the label, 3.3% of people on canagliflozin and 1.5% on placebo had an amputation in CANVAS, and 1.5% against 0.9% in CANVAS-R [24]. FDA required a boxed warning in 2017 and removed it in 2020, stating that the risk, while still increased with canagliflozin, is lower than previously described [27]. In the canagliflozin kidney trial, amputation rates did not differ significantly (12.3 against 11.2 events per 1,000 patient-years) [8][24]. Pooling seven randomized trials of the class, the association was not significant with a random effects model (RR 1.28, 95% CI 0.93 to 1.76), was significant for canagliflozin alone (RR 1.59, 95% CI 1.26 to 2.01), and was not significant for dapagliflozin or empagliflozin [28]. A network meta-analysis found raised odds for the class (OR 1.2, 95% CI 1.1 to 1.3) and for canagliflozin (OR 1.6, 95% CI 1.1 to 2.4) [29]. Across four empagliflozin outcome trials the rate was 5.0 against 4.3 events per 1,000 patient-years (HR 1.05, 95% CI 0.81 to 1.36) [20].
Population: Adults with type 2 diabetes and raised cardiovascular risk followed an average of 5.7 years in CANVAS and 2.1 years in CANVAS-R [13][24]; 4,401 adults in the kidney trial [8]; 42,453 participants in seven trials with events [28]; 96,522 participants in 36 studies [29].
Sources: [13], [24], [27], [8], [28], [29], [20]
Caveats: The band above is the share with an amputation on canagliflozin in CANVAS. There is no boxed warning on the current labels. Heterogeneity between trials was substantial (I² 62%) [28].

Testosterone therapyModerateStudied in menCommon

Fractures

Shown to raise clinical fractures, 3.5% against 2.5% on placebo, in men with low testosterone. Major osteoporotic, spine and hip fractures did not differ.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: After a median of 3.19 years, a clinical fracture occurred in 91 men (3.50%) on testosterone and 64 (2.46%) on placebo (HR 1.43, 95% CI 1.04 to 1.97). Fracture incidence also appeared higher on testosterone for every other fracture end point [4]. A meta-analysis of 2 RCTs found more clinical fractures on testosterone (RR 1.55, 95% CI 1.21 to 1.97), with no difference in major osteoporotic fractures (RR 0.62, 95% CI 0.12 to 3.35), vertebral fractures (RR 0.87, 95% CI 0.29 to 2.58) or hip fractures (RR 1.02, 95% CI 0.33 to 3.09) [13].
Population: 5,204 men aged 45 to 80 with low testosterone in the TRAVERSE fracture subtrial [4]; 2,711 hypogonadal men in 2 RCTs [13].
Sources: [4], [13]
Caveats: The meta-analysis pooled only 2 trials, and the added fractures were mainly non-major ones [13]. The TRAVERSE subtrial was not designed around men with osteoporosis.

YogaModerateCommon

Adverse events in trials, mostly more back pain

Shown to raise adverse events, mainly more back pain, against no exercise in back pain trials, at about 4.3% against 0.9%. Rates matched other exercise and usual care.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: In back pain trials, adverse events at 6 to 12 months occurred in 43 per 1,000 with yoga and 9 per 1,000 with no exercise (risk ratio 4.76, 95% CI 2.08 to 10.89, low certainty), and in 84 per 1,000 with yoga and 91 per 1,000 with other exercise [1]. Across 94 trials with 8,430 participants, intervention-related, nonserious and serious adverse events did not differ between yoga and usual care or exercise. Against psychological or educational programs, yoga had more intervention-related (odds ratio 4.21) and nonserious adverse events (odds ratio 7.30), with similar serious events [15]. In the back pain trial against physical therapy, adverse events were mostly mild self-limited joint and back pain, as frequent as with physical therapy [3].
Population: 1,037 participants in 8 back pain trials [1]. 8,430 participants in 94 trials [15]. 320 adults [3].
Sources: [1], [15], [3]
Caveats: Only 94 of 301 trials reported on adverse events [15].

59 more

YogaModerateCommon

Injuries reported by people who practice

Linked to injuries or other adverse events in 4.6% of practitioners over 12 months in observational studies, mostly sprains and strains, with 1.9% serious.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: Across 9 observational studies, 22.7% of practitioners reported an adverse event during a yoga class, 12-month prevalence was 4.6% (95% CI 3.8% to 5.4%) and lifetime prevalence ranged from 21.3% to 61.8%. Serious adverse events occurred in 1.9%. Against people who did not practice, falls (odds ratio 0.90) and fall injuries (odds ratio 1.04) did not differ, and meniscus injuries were more common (odds ratio 1.72, 95% CI 1.23 to 2.41) [16].
Population: 9,129 yoga practitioners and 9,903 non-practitioners in 9 studies [16].
Sources: [16]
Caveats: Observational and self-reported. People who choose yoga may differ from those who do not.

ZolpidemModerateCommon

Falls and fractures in older adults

Linked to more fractures in older adults in three meta-analyses of observational studies, with hip fracture risk about 1.9 to 2.8 times higher.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: Z-drugs were associated with more fractures (OR 1.63, 95% CI 1.42 to 1.87) and zolpidem with more injuries (OR 2.05, 95% CI 1.95 to 2.15). The odds of falls trended higher without reaching significance (OR 2.40, 95% CI 0.92 to 6.27) [22]. Z-drug use was associated with hip fracture (RR 1.90, 95% CI 1.68 to 2.13), about as strongly as benzodiazepines (RR 1.52, 95% CI 1.37 to 1.68), and most strongly in short-term use (RR 2.39, 95% CI 1.74 to 3.29) [23]. Zolpidem use was associated with fracture at any site (RR 1.84, 95% CI 1.67 to 2.03) and hip fracture (RR 2.80, 95% CI 2.19 to 3.58) [24]. In people with dementia, higher-dose Z-drugs were associated with fractures (HR 1.67, 95% CI 1.13 to 2.46) and falls (HR 1.33, 95% CI 1.06 to 1.66) against non-users with sleep disturbance [25]. In non-US zolpidem trials, 30 of 1,959 people (1.5%) reported falls, 28 of them aged 70 or older, and 23 of those 28 were on doses above 10 mg [13].
Population: 14 observational studies, 830,877 people in the fracture analysis [22]; observational studies of hip fracture in older people [23]; 1,092,925 participants in 9 studies [24]; 3,532 people with dementia newly prescribed Z-drugs [25]; 1,959 people in non-US trials [13].
Sources: [22], [23], [24], [25], [13]
Caveats: Observational and open to confounding, with considerable heterogeneity in the fracture and falls analyses (I² 90% and 95%) [22]. The band above is the fall rate in trials, which had no placebo figure beside it. Minimal or inconsistent excess risk was seen at lower doses in people with dementia [25].

AllopurinolLimitedCommon

Gout flares when starting

The label warns that gout flares can occur when allopurinol is started, even as uric acid falls, and lists acute gout attacks at 1% or more.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: The label states that flares have been reported during initiation, even when normal or low serum uric acid has been reached, because urate is mobilized from tissue deposits, and that it may take several months to control flares. It lists acute attacks of gout among reactions occurring in 1% or more [5]. In a year-long trial that gave naproxen or colchicine for the first 8 weeks, 64% on allopurinol still had a flare during weeks 9 to 52 [2].
Population: People with gout starting allopurinol, as described in the label [5]; 760 people with gout [2].
Sources: [5], [2]
Caveats: The trial had no placebo group, and people with gout have flares without treatment. The band here is the label's 1% or more.

IsotretinoinLimitedStudied in childrenCommon

Bone density and growth in teenagers

One trial found bone density declines in 9% of teenagers during a 20-week course, with no recovery 2 to 3 months later. Bone age changed in 3%.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: In a trial of patients aged 12 and older, 27 of 306 (9%) had declines in bone mineral density of 4% or more at the lumbar spine or total hip, or 5% or more at the femoral neck, during 20 weeks. Repeat scans 2 to 3 months later showed no recovery. Bone age changed in 9 of 289 (3%). The label reports cases of premature growth plate closure, fractures and hyperostosis [3].
Population: 306 patients aged 12 and older with bone scans, and 289 with hand x-rays [3].
Sources: [3]
Caveats: No untreated comparison group, and long-term follow-up was in 7 patients [3].

BisphosphonatesModerateUncommon

Atypical thigh bone fractures

Linked to rare atypical thigh bone fractures that rose with years of use and fell quickly after stopping. The label warns of them.

The numbers and sources

How common: Uncommon: between 1 in 1,000 and 1 in 100 people.
Effect: Bisphosphonate use was linked to subtrochanteric, femoral shaft and atypical femur fractures (adjusted RR 1.70, 95% CI 1.22 to 2.37), and 1.62 with at least 5 years of use [16]. Among 196,129 women on bisphosphonates, 277 had an atypical femur fracture. Against under 3 months of use, the hazard ratio was 8.86 at 3 to 5 years and 43.51 at 8 years or more. Asian women had 4.84 times the risk of White women. Over 3 years in White women, 149 hip fractures were prevented for every 2 atypical fractures, and in Asian women 91 for 8 [17]. The alendronate label warns of atypical low-energy femur fractures [15].
Population: 11 observational studies [16]; 196,129 women aged 50 or older followed up to 10 years [17].
Sources: [16], [17], [15]
Caveats: Observational, with high heterogeneity in [16]. The 277 cases in [17] were counted over a decade of follow-up.

YogaLimitedRare

Emergency department visits for yoga injuries

Linked to 29,590 US emergency department visits from 2001 to 2014, mostly trunk sprains and strains, at the highest rate in people aged 65 and over.

The numbers and sources

How common: Rare: fewer than 1 in 1,000 people.
Effect: Of 29,590 yoga-related injuries seen in US emergency departments, 46.6% involved the trunk and 45.0% were sprains or strains. In 2014, the injury rate was 57.9 per 100,000 participants aged 65 and over, 17.7 at 45 to 64 and 11.9 at 18 to 44 [17].
Population: National Electronic Injury Surveillance System data, with participant numbers from national health surveys [17].
Sources: [17]
Caveats: Counts only injuries that reached an emergency department, and the participant denominators are survey estimates [17].

BenzodiazepinesStrongHow often is not known

Falls and fractures

Linked to about 1.4 to 1.6 times the odds of falls in older adults and 1.5 times the risk of hip fracture in observational studies.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In older adults, benzodiazepine use was associated with falls at an odds ratio of 1.57 (95% CrI 1.43 to 1.72), or 1.41 (95% CrI 1.20 to 1.71) in the adjusted estimate [25]. A later meta-analysis found an odds ratio of 1.42 (95% CI 1.22 to 1.65), with 1.81 for long-acting and 1.27 for short-acting drugs [26]. For hip fracture, the relative risk was 1.52 (95% CI 1.37 to 1.68), and 2.40 (95% CI 1.88 to 3.05) with short-term use [27].
Population: 79,081 older participants in 22 studies of 9 drug classes [25]; 248 studies of psychotropic drugs and falls [26]; studies of benzodiazepines and hip fracture in older people [27].
Sources: [25], [26], [27]
Caveats: Observational, with substantial heterogeneity. Prescribing bias could not be accounted for [26]. The sources give no absolute rates.

Growth hormoneStrongHow often is not known

Swelling, joint pain and carpal tunnel syndrome

Shown to raise swelling, joint pain and carpal tunnel symptoms, with swelling about 6 times as likely in adults with deficiency.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Healthy older adults on growth hormone were more likely to have soft tissue swelling, joint pain, carpal tunnel syndrome and breast enlargement in men [1]. Young athletes had more swelling and fatigue [2]. In adults with deficiency, the relative risk of swelling was 6.07 (95% CI 4.34 to 8.48) and of joint stiffness 4.17 (95% CI 1.4 to 12.38) [6]. In orthopedic studies, the odds of swelling of the limbs were 3.03, of carpal tunnel symptoms 3.85 and of joint pain 2.94 on growth hormone [5].
Population: As in [1], [2], [6] and [5].
Sources: [1], [2], [6], [5]
Caveats: [1] and [2] report these as more frequent without pooled rates in the abstract.

Proton pump inhibitorsStrongHow often is not known

Fractures

Linked to about 1.3 times the hip fracture risk in observational studies. A 3-year randomized trial found no difference in fractures. The label warns of fracture risk with long-term use.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: One meta-analysis of observational studies found higher risk of hip (RR 1.30, 95% CI 1.19 to 1.43), spine (RR 1.56) and any-site fractures (RR 1.16), and no significant rise with H2 blockers (RR 1.12, 95% CI 0.97 to 1.30) [22]. A second found hip fracture RR 1.26 (95% CI 1.16 to 1.36) [23]. In a randomized trial, fractures did not differ from placebo over a median 3 years [19]. The label states the risk was higher with multiple daily doses and use for a year or longer [3].
Population: 11 observational studies, mostly of older adults [22]; 18 studies with 244,109 fracture cases [23]; 17,598 adults in a randomized trial [19].
Sources: [22], [23], [19], [3]
Caveats: Residual confounding cannot be excluded in observational studies [22], and heterogeneity was high (I² 71.9%) [23]. Three years may be too short for a trial to detect a change in fracture risk.

SSRIsStrongHow often is not known

Fractures

Linked to about 1.6 times the fracture risk in observational studies, a figure that fell to about 1.4 in studies adjusting for more risk factors.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Across 37 observational studies, SSRI use was associated with fracture (RR 1.62, 95% CI 1.52 to 1.73) [50]. An earlier meta-analysis of 12 studies found an odds ratio of 1.69 (95% CI 1.51 to 1.90), which was 1.38 (95% CI 1.27 to 1.49) in studies adjusting for four or more fracture risk factors, and 2.06 for hip or femur fracture [51].
Population: 14 case-control and 23 cohort studies [50]; 7 case-control and 5 cohort studies [51].
Sources: [50], [51]
Caveats: Observational with very high heterogeneity (I² about 90%). Depression itself is tied to falls and fracture, and studies adjusting for bone density showed a smaller association [50]. The sources give no absolute rate.

Vegan dietStrongHow often is not known

Low calcium intake

Linked to calcium intake below recommendations in most vegans studied, the lowest of any diet group.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Calcium intake in most vegans was below the 750 mg a day recommendation [7]. A review of 141 studies found vegans had the lowest calcium intake [9].
Population: Vegans in 48 European studies [7]; adults in 141 studies [9]. Observational.
Sources: [7], [9]
Caveats: Intake, not a blood measure. Calcium intake partly explained the vegan fracture finding in one cohort [12].

ADHD stimulantsModerateHow often is not known

Slower growth in children

Shown to slow growth in children, with about 2 cm less height and 2.7 kg less weight over 3 years of daily methylphenidate.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The methylphenidate label reports that children aged 7 to 13 treated 7 days a week for 14 to 36 months grew on average about 2 cm less in height and 2.7 kg less in weight over 3 years, with no growth rebound during that period [21]. In the long-term follow-up of that trial, people with consistent or inconsistent stimulant use through childhood were 2.55 cm shorter as adults than those with negligible use [29]. In a 4-week trial in adolescents, mean weight change was -1.1 lb on 10 mg and -2.8 lb on 20 mg of extended-release mixed amphetamine salts [17].
Population: Children aged 7 to 13 followed for up to 3 years [21]; the same trial's participants followed to about age 25, with 289 classmates as a comparison group [29]; adolescents in a 4-week trial [17].
Sources: [21], [29], [17]
Caveats: The sources report average differences, not the share of children affected. The adult height comparison is between self-selected patterns of use, not randomized groups [29].

Chiropractic and spinal manipulationModerateHow often is not known

Serious adverse events overall

Serious events are rare in published estimates, from 1 per 2 million manipulations to 13 per 10,000 patients, and none was reported in trials of acute back pain.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: A scoping review of 250 articles found serious adverse event estimates ranging from 1 per 2 million manipulations to 13 per 10,000 patients, including spinal or neurological problems and cervical artery strokes [23]. No trial of acute low back pain reported a serious adverse event [6]. In older adults, stroke of any type within 7 days was 1.2 per 1,000 after chiropractic visits and 1.4 per 1,000 after primary care visits for neck pain (HR 0.39), and slightly higher at 30 days (HR 1.10) [24]. In one chronic back pain trial, a safety board judged one serious adverse event possibly related to manipulation [3].
Population: 250 articles [23]. 26 trials [6]. 1,157,475 Medicare beneficiaries aged 66 to 99 [24]. 47 trials [3].
Sources: [23], [6], [24], [3]
Caveats: The estimates span two frequency bands and come from different designs. The reviewers write that incidence and cause are hard to establish [23].

Inhaled corticosteroidsModerateStudied in childrenHow often is not known

Slower growth in children

Shown to slow children's growth by about 0.48 cm in the first year, with a smaller effect after. Children on budesonide were 1.2 cm shorter as adults.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Against placebo or non-steroid drugs, inhaled steroids lowered growth velocity by 0.48 cm a year (95% CI -0.65 to -0.30) in the first year. In the second year the difference was not significant (MD -0.19 cm a year) [18]. In the early intervention trial, children under 11 on budesonide grew 1.34 cm less over 3 years [3]. In a trial following children into adulthood, adult height was 1.2 cm lower on budesonide than placebo (95% CI -1.9 to -0.5), similar to the gap after 2 years, and higher doses in the first 2 years were linked to lower adult height [14].
Population: 8,471 children with mild to moderate persistent asthma in 25 trials [18]; children aged 5 to 13 at the start, 943 measured as adults [14].
Sources: [18], [3], [14]
Caveats: The effect differed by drug [18]. The loss in adult height was not progressive or cumulative [14].

Inhaled corticosteroidsModerateHow often is not known

Fractures

Mixed results: one meta-analysis linked inhaled steroids to about 1.27 times the fracture risk in COPD trials, while a Cochrane review found no major effect on fractures over three years.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In COPD trials, fluticasone and budesonide were associated with fractures (Peto OR 1.27, 95% CI 1.01 to 1.58), and in observational studies with OR 1.21, rising 9% per 500 μg of beclomethasone-equivalent dose [19]. A Cochrane review found long-term studies that measured bone effects generally showed no major effect on fractures or bone density over three years [9].
Population: 17,513 participants in 16 trials and 69,000 in 7 observational studies [19]; long-term COPD trials [9].
Sources: [19], [9]
Caveats: The pooled trial estimate is just significant, and most trials were of fluticasone [19].

LevothyroxineModerateHow often is not known

Fractures and bone loss

Linked to more fractures in adults 70 or older, rising with dose, and in people on levothyroxine whose TSH was suppressed.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In adults aged 70 or older, current levothyroxine use was associated with fracture (adjusted OR 1.88, 95% CI 1.71 to 2.05) against remote use. Against low cumulative doses (under 0.044 mg a day), medium doses had an OR of 2.62 and high doses (over 0.093 mg a day) 3.45 [13]. A suppressed TSH was associated with fracture (HR 2.02) [11]. The label states over-replacement can increase bone resorption and decrease bone mineral density [10]. In a meta-analysis, people with overt hypothyroidism on levothyroxine had lower lumbar spine bone density than healthy controls (SMD -0.28, 95% CI -0.55 to -0.02). Hip sites did not differ, and no difference was found in subclinical hypothyroidism [14].
Population: 213,511 levothyroxine users aged 70 or older in Ontario, 22,236 with a fracture over a mean 3.8 years [13]; 17,684 people in Scotland [11]; 13 studies [14].
Sources: [13], [11], [10], [14]
Caveats: [13] and [11] are observational. [14] compared treated patients with healthy people, not with untreated patients. The dose relation in [13] may partly reflect who receives higher doses.

Massage therapyModerateHow often is not known

More pain or soreness after treatment

Shown in trials as the most common side effect, with increased pain reported in 1.5% to 25% of participants across back pain trials.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In low back pain trials, increased pain intensity was the most common adverse event, in 1.5% to 25% of participants [4]. In neck pain trials, the 4 that reported side effects all reported pain after treatment, and one found low blood pressure after treatment in 22% of participants [6].
Population: 25 low back pain trials [4]. 15 neck pain trials, 4 reporting side effects [6].
Sources: [4], [6]
Caveats: The range spans two frequency bands and no pooled rate was given. Most trials did not report side effects [6].

Opioid pain relieversModerateHow often is not known

Fractures

Linked to fractures, with a pooled relative risk of 1.88 across 8 cohort studies.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Across 8 cohort studies, opioid use was associated with fracture (RR 1.88, 95% CI 1.51 to 2.34) [26]. A federal evidence review also found opioids associated with fractures, falls, heart attack, overdose and death from any cause in observational studies, with a dose response for all but fractures and falls [7].
Population: 8 cohort studies, 6 rated high quality [26]; 40 observational studies [7].
Sources: [26], [7]
Caveats: Observational. People prescribed opioids differ from those who are not in ways that also affect falls and fractures.

Oral corticosteroidsModerateHow often is not known

Fractures and bone loss

Linked to fractures that rise with dose, with hip fracture 1.61 times and spine fracture 2.60 times as frequent as in matched non-users.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Against matched controls, the relative rate during oral corticosteroid treatment was 1.33 for non-spine fracture (95% CI 1.29 to 1.38), 1.61 for hip fracture (1.47 to 1.76) and 2.60 for spine fracture (2.31 to 2.92). Hip fracture risk was 0.99 below 2.5 mg of prednisolone a day, 1.77 at 2.5 to 7.5 mg and 2.27 at 7.5 mg or more, and spine fracture risk was 1.55, 2.59 and 5.18. Risk fell toward baseline after treatment stopped [17]. In the 30 days after starting a short course, the fracture rate was higher (rate ratio 1.87, 95% CI 1.69 to 2.07) [22]. The label states that corticosteroids decrease bone formation and increase bone resorption and may lead to osteoporosis at any age [20].
Population: 244,235 adult oral corticosteroid users and 244,235 matched controls in UK general practice [17]; 327,452 US adults given a short course [22].
Sources: [17], [22], [20]
Caveats: Observational. The illnesses treated with corticosteroids may themselves affect bone.

Ozone therapyModerateHow often is not known

Minor effects in trials

Reviews of knee osteoarthritis trials reported a low prevalence of minor adverse events, and one placebo trial recorded only puncture accidents.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Seven of 8 systematic reviews of ozone for knee osteoarthritis reported a low prevalence of minor adverse events [7]. In a 98-person placebo-controlled trial, adverse events occurred in 3 people, 1 given ozone and 2 given placebo, and all were puncture accidents [6]. An umbrella review found safety outcomes reported too inconsistently across conditions to draw firm conclusions [12].
Population: 8 reviews of knee osteoarthritis trials [7]. 98 adults [6]. 7 meta-analyses [12].
Sources: [7], [6], [12]
Caveats: Rates were not given in the review abstracts.

Physical activityModerateHow often is not known

Minor adverse events in exercise trials

Shown to raise non-serious adverse events by about a fifth in exercise therapy trials, one extra event for every 6 people, with no rise in serious events.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Exercise therapy raised non-serious adverse events (RR 1.19, 95% CI 1.09 to 1.30), a number needed to harm of 6. Serious adverse events did not rise (RR 0.96, 95% CI 0.90 to 1.02) [21].
Population: 38,517 participants in 375 trials for non-serious and 38,368 in 378 trials for serious events, drawn from 180 Cochrane reviews [21].
Sources: [21]
Caveats: Trials were of exercise therapy, often in people with a health condition, and recorded adverse events in different ways. The abstract does not name the events.

Progestins for endometriosis and heavy bleedingModerateStudied in womenHow often is not known

Bone loss with the medroxyprogesterone injection

FDA states, in a boxed warning, that women using the injection may lose significant bone mineral density, more with longer use, and that the loss may not fully reverse.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: On Depo-SubQ Provera 104, spine bone density fell 2.7% (95% CI 2.3 to 3.1) at 1 year and 4.1% (95% CI 3.5 to 4.6) at 2 years, and total hip density 1.7% and 3.5%. After 5 years on the intramuscular injection, spine density was 5.38% lower against 0.43% higher in women using no hormonal contraception, and 2 years after stopping it was still 3.13% lower. In adolescents treated for more than 2 years, hip and femoral neck density had not recovered 5 years after stopping [1]. In 6-month endometriosis trials, bone loss was smaller than with leuprolide and returned to baseline within 12 months of stopping [10][11].
Population: 166 women at 1 year and 106 at 2 years [1]. 33 women on the injection and 105 controls at 5 years, 12 and 60 at 7 years [1]. 274 and 300 women with endometriosis [10][11].
Sources: [1], [10], [11]
Caveats: The label states it is unknown whether use in adolescence lowers peak bone mass or raises later fracture risk [1]. Long-term follow-up groups were small.

PRP injectionsModerateHow often is not known

Frozen shoulder after rotator cuff injections

Pooled trials found overall complications similar to other shoulder injections, and frozen shoulder more often after PRP.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Across 19 trials, complications occurred in 9.41% given PRP and 10.13% given comparison injections (P = .67), and frozen shoulder was significantly more frequent after PRP (P = .01). The most common complications were persistent pain, frozen shoulder, worsening of the tear and bursitis. No serious events were reported [18].
Population: 1,043 patients with rotator cuff problems in 19 trials, 510 given PRP [18].
Sources: [18]
Caveats: The abstract does not give the frozen shoulder rate.

SGLT2 inhibitorsModerateHow often is not known

Bone fractures

Mixed results: one canagliflozin trial found more fractures than placebo, while a later canagliflozin trial found no difference and a pooled analysis found a small rise with dapagliflozin.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In the CANVAS trial, adjudicated fractures occurred at 1.59 and 1.79 events per 100 patient-years on canagliflozin 100 mg and 300 mg against 1.09 on placebo. The imbalance appeared within 26 weeks, and fractures were more often low trauma [24]. In the canagliflozin kidney trial, fracture rates did not differ significantly from placebo [8]. In a network meta-analysis of 66 studies, dapagliflozin was associated with fracture (OR 1.1, 95% CI 1.0 to 1.2) [29].
Population: Adults with type 2 diabetes and raised cardiovascular risk [24]; 4,401 adults with type 2 diabetes and kidney disease [8]; 102,862 participants in 66 studies [29].
Sources: [24], [8], [29]
Caveats: The sources give rates per patient-years, not the share of people.

Strength trainingModerateHow often is not known

Muscle and joint complaints in trials

Reported in trials: mostly minor muscle and joint complaints such as joint pain and soreness, with serious adverse events rare and poorly recorded.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Across 121 trials in older adults, adverse events were poorly recorded. Joint pain and muscle soreness were reported in many trials that monitored them, serious adverse events were rare, and none was reported as directly related to training [9]. In falls prevention trials, reported events were mainly non-serious and musculoskeletal, with two serious events, a pelvic stress fracture and hernia surgery, in one trial [16]. The high-intensity bone trial recorded one minor back spasm [15].
Population: 6,700 older adults in 121 trials [9]. 6,019 participants in 27 trials that reported adverse events [16]. 101 postmenopausal women [15].
Sources: [9], [16], [15]
Caveats: Most trials did not monitor adverse events closely in both groups, so rates cannot be given [9][16].

TesamorelinModerateHow often is not known

Joint pain, muscle pain and swelling

Shown to raise joint pain, muscle pain and leg swelling slightly in adults with HIV, such as muscle pain in 6% against 2% on placebo.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Over 26 weeks, joint pain occurred in 13% on tesamorelin and 11% on placebo, pain in a limb in 6% against 5%, muscle pain in 6% against 2%, swelling of the legs in 6% against 2%, and carpal tunnel syndrome in 1% against 0% [6]. A meta-analysis of 5 RCTs also linked tesamorelin with joint pain, muscle pain and tingling (paresthesia) [5].
Population: 543 adults on tesamorelin and 263 on placebo [6]; 5 RCTs [5].
Sources: [6], [5]
Caveats: The label groups these as fluid retention effects [6].

Thiazide diureticsModerateHow often is not known

Gout and high uric acid

Linked to more than twice the risk of new gout with diuretic use in people with high blood pressure. Trials found thiazides raised uric acid.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In a UK case-control study, current diuretic use was associated with incident gout among people with hypertension (RR 2.36, 95% CI 2.21 to 2.52) [19]. Across placebo-controlled trials, thiazides raised uric acid in a dose-related way [1]. Gout was more frequent on hydrochlorothiazide than placebo in the kidney stone trial [13]. The label states that hyperuricemia may occur and acute gout may be precipitated [14].
Population: 24,768 people with new gout and 50,000 matched controls [19]; 11,282 people in 60 trials [1].
Sources: [19], [1], [13], [14]
Caveats: The case-control study grouped all diuretics together and gives no absolute rate [19].

Vegan dietModerateHow often is not known

Fractures and bone density

Linked to about 2.3 times the hip fractures and 43% more total fractures in vegans than meat eaters in a UK cohort, and to lower bone density.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In EPIC-Oxford over 17.6 years, vegans had more hip fractures (HR 2.31, 95% CI 1.66 to 3.22), about 14.9 more per 1,000 people over 10 years, and more total (HR 1.43), leg (HR 2.05) and other main site fractures than meat eaters [16]. An earlier analysis found 30% more fractures in vegans (rate ratio 1.30, 95% CI 1.02 to 1.66), which fell to 1.15 (95% CI 0.89 to 1.49) after adjusting for energy and calcium intake, and to 1.00 in people eating at least 525 mg of calcium a day [12]. Across 20 studies, vegans had lower bone mineral density at the femoral neck and lumbar spine and higher fracture rates than omnivores [17].
Population: 54,898 UK adults including 1,982 vegans [16]; 34,696 UK adults including 1,126 vegans [12]; 37,134 people in 20 studies [17]. Observational.
Sources: [12], [16], [17]
Caveats: Observational. In the later analysis, the links remained after adjusting for calcium and protein intake [16], while in the earlier analysis they did not [12]. Wrist and ankle fractures did not differ [16].

Vegetarian dietModerateHow often is not known

Lower bone density and hip fracture

Linked to lower bone density and about 25% more hip fractures in vegetarians than meat eaters, though an earlier analysis of the same cohort found no difference.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Across 20 studies (37,134 participants), vegetarians and vegans had lower bone mineral density at the femoral neck and lumbar spine than omnivores [20]. In EPIC-Oxford over 17.6 years, vegetarians had more hip fractures than meat eaters (HR 1.25, 95% CI 1.04 to 1.50), about 2.9 more per 1,000 people over 10 years [21]. An earlier EPIC-Oxford analysis over 5.2 years found no difference in total fractures for vegetarians (rate ratio 1.00, 95% CI 0.89 to 1.13) [22].
Population: Adults in 20 observational studies [20]; 54,898 UK adults, 15,499 vegetarians [21]; 34,696 UK adults [22].
Sources: [20], [21], [22]
Caveats: Observational. The earlier analysis used self-reported fractures [22]. The hip fracture link stayed significant after adjusting for calcium and protein intake [21].

Alpha-blockersLimitedStudied in menHow often is not known

Falls and fractures in older men

Linked to slightly more falls and fractures in the 90 days after older men started tamsulosin, alfuzosin or silodosin, with absolute increases under 0.2%.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Men starting a prostate-selective alpha-blocker had more falls (OR 1.14, 95% CI 1.07 to 1.21, absolute increase 0.17%) and fractures (OR 1.16, 95% CI 1.04 to 1.29, absolute increase 0.06%) than matched men who did not, and more hypotension (OR 1.80) and head trauma (OR 1.15) [24].
Population: 147,084 men aged 66 or older in Ontario and as many matched men, 84% on tamsulosin [24].
Sources: [24]
Caveats: Observational, with possible unmeasured confounding such as mobility. The authors note the absolute risks are probably underestimates [24].

Atomoxetine and viloxazineLimitedStudied in childrenHow often is not known

Growth in children

The label reports that weight and height gain in children on atomoxetine lagged behind population norms in long-term open-label studies.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In long-term open-label studies, weight and height gain in children treated with atomoxetine lagged behind that predicted from population data. Weight decreased in 7% of poor CYP2D6 metabolizers against 4% of others [9].
Population: Children in long-term open-label studies [9].
Sources: [9]
Caveats: Compared with population norms, not a randomized control group.

BisphosphonatesLimitedHow often is not known

Bone, joint and muscle pain

The label warns of severe, sometimes incapacitating bone, joint or muscle pain reported after approval, while placebo-controlled alendronate trials found similar rates.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The alendronate label reports postmarketing cases of severe bone, joint or muscle pain beginning a day to months after starting, mostly relieved after stopping, with some recurrence on rechallenge. In its placebo-controlled trials, the share with these symptoms was similar on alendronate and placebo [15]. In the main zoledronic acid trial, bone pain was 5.8% against 2.3% on placebo [13].
Population: Postmarketing reports and trial populations [15][13].
Sources: [15], [13]
Caveats: Postmarketing reports cannot give a rate.

BupropionLimitedHow often is not known

Falls in older adults

One trial found the rate of falls was highest when bupropion was added to an existing antidepressant in adults aged 60 and older.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Among older adults with treatment-resistant depression assigned to added aripiprazole, added bupropion or a switch to bupropion, the rate of falls was highest with added bupropion [11].
Population: 619 adults aged 60 and older, 206 of them assigned to added bupropion [11].
Sources: [11]
Caveats: One trial with no placebo arm. The abstract gives no fall rates.

Chiropractic and spinal manipulationLimitedStudied in childrenHow often is not known

Harms in infants and children

Case reports: 9 serious events, including brain bleeding and paralysis, and 20 delayed diagnoses were reported after spinal manipulation in children.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: A review of 13 studies found 14 direct adverse events in children, 9 of them serious, including subarachnoid hemorrhage and paraplegia. Another 20 cases involved delayed diagnosis of conditions such as diabetes, neuroblastoma and meningitis [25]. A later review found 5 case reports of severe harm after high-velocity thrusts in 4 infants and 1 child, and mild, transient harms after gentle mobilizations [12].
Population: Children aged 18 or younger in 13 studies [25]. 26 studies in infants, children and adolescents [12].
Sources: [25], [12]
Caveats: Neither causation nor incidence can be drawn from these reports [25]. The later reviewers write that severe harms were scarce, poorly described and likely tied to missed underlying disease [12].

CoffeeLimitedHow often is not known

Fracture

Mixed results: an umbrella review linked coffee to fracture in women but not men, while a meta-analysis found no link with hip fracture.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The umbrella review reported an association between coffee and fracture risk in women and not in men [2]. Pooling 14 studies, hip fracture risk at the highest against the lowest coffee intake was 0.94 (95% CI 0.71 to 1.17) [24].
Population: Meta-analyses of observational studies [2]. 195,992 people and 9,958 hip fractures in 6 cohort and 8 case-control studies [24].
Sources: [2], [24]
Caveats: Observational. The umbrella review abstract gives no effect size for women [2].

DairyLimitedHow often is not known

Death and fractures in two Swedish cohorts

Linked in two Swedish cohorts to 15% higher death per daily glass of milk in women and 3% in men, and to more hip fractures in women.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In women, 3 or more glasses of milk a day against less than one was linked to higher death from any cause (HR 1.93, 95% CI 1.80 to 2.06). Each daily glass was linked to higher death in women (HR 1.15, 95% CI 1.13 to 1.17) and men (HR 1.03, 95% CI 1.01 to 1.04). In women, each glass was linked to hip fracture (HR 1.09, 95% CI 1.05 to 1.13) and not to any fracture (HR 1.02). In men, fracture was not linked [17].
Population: 61,433 Swedish women followed a mean 20.1 years and 45,339 Swedish men followed 11.2 years [17].
Sources: [17]
Caveats: One pair of cohorts. The authors call for cautious interpretation because of residual confounding and reverse causation [17]. Pooled cohorts found no link between milk and death [11].

Direct oral anticoagulants (DOACs)LimitedHow often is not known

FDA boxed warning: bleeding around the spine

FDA states that epidural or spinal hematomas may occur with spinal anesthesia or spinal puncture in people on these drugs, and may cause long-term or permanent paralysis.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The boxed warning on all four labels describes epidural or spinal hematomas in people receiving neuraxial anesthesia or undergoing spinal puncture, which may result in long-term or permanent paralysis. The listed risk factors include indwelling epidural catheters and use of other drugs that affect clotting, such as NSAIDs, platelet inhibitors and other anticoagulants [22][23][20][21].
Population: Boxed warnings in labeling, with no study counts [22][23].
Sources: [22], [23], [20], [21]
Caveats: The labels give no rate.

FluorideLimitedHow often is not known

Skeletal fluorosis

Limited evidence: high fluoride exposure was linked to about twice the skeletal fluorosis, with risk rising almost linearly up to 5 mg/L in water.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Across 23 studies, skeletal fluorosis was twice as common at the highest against the lowest fluoride exposure (RR 2.05, 95% CI 1.60 to 2.64), and 2.73 times as common for drinking water (95% CI 1.92 to 3.90). Risk rose almost linearly up to 5 mg/L in water, and moderate to severe forms rose at 5.00 mg/L [17]. A 53-year-old woman with a metatarsal fracture had a spine bone density T-score of +11, with very high fluoride levels linked to excessive tea and toothpaste consumption [18].
Population: Observational studies, many in regions where fluorosis is endemic [17]. One case report [18].
Sources: [17], [18]
Caveats: Observational data at moderate or high risk of bias [17].

GabapentinoidsLimitedHow often is not known

Accidental overdose, injuries and road incidents

Linked to more unintentional overdoses, injuries and road traffic incidents during treatment periods in one Swedish cohort, driven by pregabalin.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Comparing periods on and off treatment in the same people, gabapentinoid treatment was associated with unintentional overdose (HR 1.24, 95% CI 1.19 to 1.28), head or body injuries (HR 1.22, 95% CI 1.19 to 1.25) and road traffic incidents or offences (HR 1.13, 95% CI 1.06 to 1.20). Arrests for violent crime did not clearly differ (HR 1.04, 95% CI 0.98 to 1.11). Pregabalin was associated with increased hazards of every outcome, and gabapentin with decreased or no significant hazards [24].
Population: A Swedish cohort of people prescribed gabapentinoids, of whom 8.9% had an unintentional overdose and 6.3% a road traffic incident or offence during the study period [24].
Sources: [24]
Caveats: Observational. The percentages are for the whole study period, on or off treatment, so no rate on the drug can be set.

Loop diureticsLimitedHow often is not known

Fractures

Linked to more hip fractures in two observational studies, 16% more after adjustment in Denmark and about 1.5 to 1.7 times the odds in older Italians.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Ever use of loop diuretics was linked to any fracture (adjusted OR 1.04, 95% CI 1.01 to 1.07) and hip fracture (adjusted OR 1.16, 95% CI 1.10 to 1.23), with higher estimates for furosemide than bumetanide [17]. In people aged 70 to 90, current loop diuretic use was linked to hip fracture hospitalization (case-control OR 1.67, 95% CI 1.28 to 2.18, case-crossover OR 1.49, 95% CI 1.05 to 2.10) [18].
Population: 64,699 fracture cases and 194,111 controls aged 40 or older in Denmark [17]; 81,617 patients aged 70 to 90 in Lombardy, 2,153 with hip fracture [18].
Sources: [17], [18]
Caveats: Observational. The authors of [18] point to falls, and illness itself may explain part of the link.

Massage therapyLimitedHow often is not known

Serious injuries in case reports

Case reports: 138 adverse events in 40 reports from 2003 to 2013, including disc herniation, nerve injury and vertebral artery dissection, often linked to spinal manipulation within massage.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: From 2003 to 2013, 40 reports described 138 adverse events with massage, mainly disc herniation, soft tissue trauma, neurologic compromise, spinal cord injury and vertebral artery dissection. Spinal manipulation within massage was repeatedly tied to the serious events [19]. An earlier review of 16 case reports and 4 case series found most harms followed exotic techniques or massage by laypeople, rarely massage therapists. Reported events included stroke, nerve damage, pulmonary embolism and hematoma [20].
Population: 138 cases [19]. 16 case reports and 4 case series [20].
Sources: [19], [20]
Caveats: Case reports give no rate. Both reviews judged serious events rare [19][20].

MultivitaminLimitedHow often is not known

Single-nutrient harms

Limited evidence: single supplements were linked to hip fracture with vitamin A, hemorrhagic stroke with vitamin E, and kidney stones with vitamin C and calcium.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The Task Force review found limited evidence linking some single supplements to serious harms: hip fracture with vitamin A, hemorrhagic stroke with vitamin E, and kidney stones with vitamin C and calcium [4].
Population: Healthy adults in RCTs and cohorts [4].
Sources: [4]
Caveats: These are single-nutrient findings, not tests of multivitamins.

Muscle relaxantsLimitedHow often is not known

Injuries in older adults

Linked to more injuries in older adults, with 32% higher odds among recent users of a muscle relaxant in one case-control study.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Among older adults, 1.30% of those with an injury and 0.77% of matched controls had used a muscle relaxant in the prior 60 days (adjusted OR 1.32, 95% CI 1.16 to 1.50). The odds ratio was 1.42 for methocarbamol (95% CI 1.16 to 1.75) and 1.22 for cyclobenzaprine (95% CI 1.02 to 1.45) [19].
Population: 27,974 older adults with an injury and 104,303 matched controls, from a base population of 322,806 [19].
Sources: [19]
Caveats: Observational. The shares are of people with an injury who had used the drug, not of users who were injured.

Oral corticosteroidsLimitedStudied in childrenHow often is not known

Growth in children

The label warns that children treated with corticosteroids by any route may grow more slowly.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The prednisone label states that pediatric patients treated with corticosteroids by any route may experience a decrease in growth velocity, and that corticosteroids may inhibit bone growth [20].
Population: Children treated with corticosteroids, as described in the label [20].
Sources: [20]
Caveats: The label gives no rate and no size of effect.

Ozone therapyLimitedHow often is not known

Spinal infection after disc injection

Case reports: eight cases of infection of the disc and spine after oxygen-ozone therapy, one of them fatal.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: A 55-year-old woman developed L4-L5 spondylodiscitis three months after intradiscal oxygen-ozone therapy and needed surgery and IV antibiotics. A literature review identified 8 cases in all, caused by Staphylococcus aureus, streptococci, E. coli, mycobacteria and other organisms. Outcomes were favorable in all but one, who died [14].
Population: 8 published cases [14].
Sources: [14]
Caveats: Case reports give no rate.

Pet ownershipLimitedHow often is not known

Falls caused by pets

Case series: US emergency departments saw an estimated 86,629 fall injuries a year involving cats and dogs, most often fractures and bruises, highest in people 75 and older.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Dogs were involved in 7.5 times as many fall injuries as cats, and women were 2.1 times as likely as men to be injured. About 66.4% of cat-related and 31.3% of dog-related falls came from tripping over the pet, and 21.2% of dog-related falls from being pushed or pulled, as during dog walking [16].
Population: 7,456 emergency department cases, 2001 to 2006, in a nationally representative US sample [16].
Sources: [16]
Caveats: The count is not a rate among pet owners, so how often owners fall is not known [16].

Progestins for endometriosis and heavy bleedingLimitedStudied in womenHow often is not known

Fractures with the medroxyprogesterone injection

Linked to more fractures in two case-control studies of women using the contraceptive injection, with higher odds after longer use.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: One study found higher fracture odds with any use (OR 1.44, 95% CI 1.01 to 2.06) and with more than four years of use (OR 2.16, 95% CI 1.32 to 3.53). The other found higher odds with past use of one or two prescriptions (OR 1.17, 95% CI 1.07 to 1.29) and with current use of 3 to 9 prescriptions (OR 1.36) or 10 or more (OR 1.54, 95% CI 1.33 to 1.78) [16].
Population: Women in 2 case-control studies, within a review of 14 observational studies [16].
Sources: [16]
Caveats: Observational, with no meta-analysis, so women who chose the injection may differ from those who did not [16]. These were contraceptive users, not women with endometriosis.

SARMsGrey marketLimitedHow often is not known

Heart and tendon harms in case reports

Case reports: 15 men had liver injury, myocarditis or tendon rupture after SARMs. Most were athletes taking doses far above those in clinical studies.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: A systematic review of 72 articles from 2003 to 2022 found 13 case reports describing 15 people harmed after using SARMs. All were men, median age 32 (range 19 to 52), more than half were athletes, and all took SARMs by mouth for a mean of 8 weeks. Reported harms included drug-induced liver injury, myocarditis and tendon rupture. Athletes reported doses much higher than those used in clinical studies [8].
Population: 15 published cases, plus preclinical and clinical studies summarized in the review [8].
Sources: [8]
Caveats: Case reports cannot give a rate, and the review rates its own evidence as level 4. The review estimates SARM use among athletes at 1% to 3% [8].

StatinsLimitedHow often is not known

Tendon problems

Limited evidence: tendon rupture was not more common in statin users. Other observational studies linked statins to trigger finger, Achilles tendinitis and rotator cuff retears.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In a cohort of over 1 million people, native tendon rupture did not rise among statin users (HR 0.95, 95% CI 0.84 to 1.08). Other studies linked statin use to trigger finger (HR 1.435), radial styloid tenosynovitis (HR 1.365) and Achilles tendinitis (HR 1.516), and one retrospective study found more retears after rotator cuff repair (OR 6.5) [11]. Three cohort studies on shoulder tendinopathy gave conflicting results [12].
Population: 12 studies, 8 on native tendons and 4 after tendon repair [11]; 3 cohort studies [12].
Sources: [11], [12]
Caveats: Observational studies with mixed results by tendon, sex and statin type [11].

Strength trainingLimitedHow often is not known

Injuries in weight-training sports

Limited evidence: competitive weight-training sports reported about 0.24 to 7.5 injuries per 1,000 hours, mostly strains, tendinitis and sprains of the shoulder, lower back and knee.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Bodybuilding had the lowest rates, 0.24 to 1 injury per 1,000 hours. Strongman had 4.5 to 6.1 and Highland Games 7.5. The shoulder, lower back, knee, elbow and wrist or hand were most often injured [21].
Population: Competitive athletes in weightlifting, powerlifting, bodybuilding, strongman, Highland Games and CrossFit, in 20 studies [21].
Sources: [21]
Caveats: Most studies were retrospective and only 5 of 20 had a low risk of bias. These are competitive athletes, not people training for health [21].

Tai chi and qigongLimitedHow often is not known

Minor muscle and joint aches

Reported adverse events in randomized trials were mostly minor muscle and joint aches, such as knee and back pain. Two thirds of trials did not report adverse events.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Of 153 randomized trials, 50 (33%) reported adverse events and 18 (12%) described an explicit monitoring protocol. Reported events were typically minor, expected and musculoskeletal, such as knee and back pain [16].
Population: 153 randomized trials, most in older adults [16].
Sources: [16]
Caveats: Reporting was poor and inconsistent, so no rate can be given [16].

Testosterone therapyLimitedHow often is not known

Complications after knee replacement

Linked to more kidney injury, implant infection, revision surgery and fracture after knee replacement in people using testosterone.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In people having a total knee replacement, testosterone use was linked to more acute kidney injury within 3 months (OR 1.3, 95% CI 1.1 to 1.6), joint infection around the implant (OR 1.8, 95% CI 1.3 to 2.7), revision surgery (OR 1.3, 95% CI 1.2 to 1.5) and fracture around the implant (OR 2.0, 95% CI 1.1 to 3.7). Pulmonary embolism (OR 1.2, 95% CI 0.9 to 1.6) and deep vein thrombosis (OR 1.1, 95% CI 0.8 to 1.4) did not differ [18].
Population: 4 retrospective matched cohort studies [18].
Sources: [18]
Caveats: Observational studies only. They show association, not cause [18].

TrazodoneLimitedHow often is not known

Falls in older adults

Linked to falls in older adults in observational studies, at a rate similar to antipsychotics in people with dementia.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Among Medicare beneficiaries with insomnia, those receiving an insomnia medicine had more falls than matched controls (OR 2.34, 95% CI 2.31 to 2.36), with the greatest risk on benzodiazepines or trazodone [17]. In older adults with dementia, trazodone had rates of falls or major fractures similar to atypical antipsychotics (HR 0.89, 95% CI 0.73 to 1.07) and a lower rate of death (HR 0.75, 95% CI 0.66 to 0.85) [18].
Population: 1,699,913 Medicare beneficiaries, mean age 75 [17]; 6,588 older adults with dementia dispensed trazodone and 2,875 dispensed an atypical antipsychotic, 95% at low doses [18].
Sources: [17], [18]
Caveats: Observational. The authors of [17] note that use of these medicines may mark underlying illness. [18] compared trazodone with another drug, not with no treatment. Neither abstract gives an absolute fall rate for trazodone.

Vitamin ALimitedStudied in womenHow often is not known

Hip fracture

Linked to about 48% more hip fractures in postmenopausal women with the highest vitamin A intake, mainly from retinol. Beta-carotene showed no clear link.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Women in the highest fifth of total vitamin A intake, 3,000 µg retinol equivalents a day or more, had more hip fractures than those in the lowest fifth, under 1,250 µg (RR 1.48, 95% CI 1.05 to 2.07). The excess came mainly from retinol (RR 1.89, 95% CI 1.33 to 2.68, 2,000 µg or more against under 500 µg). Beta-carotene intake did not significantly change risk (RR 1.22, 95% CI 0.90 to 1.66) [12].
Population: 72,337 postmenopausal US nurses aged 34 to 77, followed 18 years, 603 hip fractures [12].
Sources: [12]
Caveats: One observational cohort. It measured intake from food and supplements, not a randomized dose.

Vitamin B6LimitedStudied in womenHow often is not known

Hip fracture

Linked to about 29% more hip fractures in postmenopausal women taking 35 mg a day or more of B6, and more with high B12 too.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Women with B6 intakes of 35 mg a day or more had more hip fractures than those under 2 mg (RR 1.29, 95% CI 1.04 to 1.59). With 35 mg or more of B6 and 20 µg or more of B12 combined, the RR was 1.47 (95% CI 1.15 to 1.89) [18].
Population: 75,864 postmenopausal US nurses followed 30 years, 2,304 hip fractures [18].
Sources: [18]
Caveats: One observational cohort with intake from food frequency questionnaires. The linear trend for B6 alone was not significant (P = 0.06) [18].

Vitamin DLimitedHow often is not known

Falls and fractures with annual mega-doses

One trial found a yearly 500,000 IU dose raised falls about 15% and fractures about 26% in women aged 70 and older.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In 2,256 women aged 70 and older, an annual oral dose of 500,000 IU raised falls (rate 83.4 versus 72.7 per 100 person-years, IRR 1.15, 95% CI 1.02 to 1.30) and fractures (IRR 1.26, 95% CI 1.00 to 1.59). The fall rate ratio was 1.31 in the first 3 months after dosing [53].
Population: Community-dwelling older women at high fracture risk, 3 to 5 years [53].
Sources: [53]
Caveats: One trial of a single large yearly dose. It does not test daily dosing.

Vitamin DLimitedHow often is not known

Lower bone density at 4,000 and 10,000 IU a day

One trial found lower forearm bone density after 3 years at 4,000 and 10,000 IU a day than at 400 IU, in adults 55 to 70. Strength did not differ.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Compared with 400 IU a day, radial volumetric bone density after 3 years was lower with 4,000 IU (-3.9 mg HA/cm³, 95% CI -6.5 to -1.3) and 10,000 IU (-7.5 mg HA/cm³, 95% CI -10.1 to -5.0). Tibial density was lower only at 10,000 IU. Bone strength did not differ [52].
Population: 311 healthy adults aged 55 to 70 without osteoporosis, Calgary [52].
Sources: [52]
Caveats: One single-center trial. Bone strength, the co-primary outcome, did not change.

YogaLimitedHow often is not known

Serious harms in case reports

Case reports: 76 published cases of harm linked to yoga, most often after headstand, shoulder stand, lotus position or forceful breathing. One person died.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Of 76 cases in 35 case reports and 2 case series, 35.5% affected the musculoskeletal system, 18.4% the nervous system and 11.8% the eyes. Ten had preconditions, mainly glaucoma and osteopenia. Fifteen recovered fully, 9 partially, 1 not at all, and 1 died [18].
Population: 76 published cases [18].
Sources: [18]
Caveats: Case reports give no rate and outcomes were not reported for every case [18].

SaunaLimited

Falls and fainting injuries

Case series: among 209 people treated for sauna injuries at one trauma center, 57.5% of injuries followed a slip or fall and 30.0% followed dizziness or fainting.

The numbers and sources

Effect: Over 17 years, 209 patients had 274 diagnoses, including bruises and sprains (41.2%), wounds (28.8%), fractures (15.3%), concussions (5.5%), burns (1.5%) and brain bleeding (1.1%). Head and face injuries mostly followed dizziness or fainting. Nine patients (4.3%) needed surgery [39].
Population: 209 patients treated at the trauma center of the Medical University of Innsbruck, 2005 to 2021 [39].
Sources: [39]
Caveats: One hospital, with no count of how many sauna visits these injuries came from.

Mixed results41

AcupunctureStrong

Knee osteoarthritis, compared with sham acupuncture

Mixed results: pooled trials found a small edge over sham acupuncture that reviewers judged below clinical relevance, and a 282-person trial found no difference from sham.

Compared with sham acupuncture [3][8][9][10]; sham laser [9]

The numbers and sources

Effect: Against sham, pain improved by 0.9 points more on a 20-point scale (SMD -0.28, 95% CI -0.45 to -0.11, 9 trials, 1,835 participants), below the reviewers' 1.3-point threshold for clinical relevance. With the shams most likely to blind participants, the benefit was smaller and not significant, and at 6 months it was borderline (SMD -0.10) [8]. Individual patient data put the edge over sham at 0.16 SD for osteoarthritis [3]. In people over 50 with chronic knee pain, needle acupuncture did not differ from sham for pain (-0.4 units on a 0 to 10 scale, 95% CI -1.2 to 0.4) or function at 12 weeks [9]. In a 480-person trial, 60.3% responded to electroacupuncture, 58.6% to manual acupuncture and 47.3% to sham at 8 weeks. Electroacupuncture differed from sham (P = 0.0234) and manual acupuncture narrowly did not (P = 0.0507) [10].
Population: 16 trials with 3,498 people with knee or hip osteoarthritis [8]. 282 adults aged 50 or older in Australia [9]. 480 adults in China [10].
Sources: [8], [3], [9], [10]
Caveats: Heterogeneity was substantial across sham-controlled trials [8]. The trials used different shams, from non-penetrating needles to inactive lasers.

Chiropractic and spinal manipulationStrong

Chronic low back pain, compared with other treatments

Shown to relieve chronic low back pain about as much as other guideline-endorsed treatments, with a small gain in function and a pain edge too small to matter clinically.

Compared with other therapies that guidelines endorse [3]; therapies that guidelines do not endorse [3]; other interventions [4]

The numbers and sources

Effect: Against other guideline-endorsed therapies, short-term pain did not differ (mean difference -3.17 on a 100-point scale, 95% CI -7.85 to 1.51) and function improved slightly (SMD -0.25, 95% CI -0.41 to -0.09). Against therapies guidelines do not endorse, pain fell 7.48 points more (95% CI -11.50 to -3.47), which the reviewers did not count as clinically better, and function improved (SMD -0.41). Results were similar at intermediate and long-term follow-up [3]. An earlier Cochrane review found a small, not clinically relevant, short-term effect on pain (mean difference -4.16, 95% CI -6.97 to -1.36) and function (SMD -0.22) against other interventions [4].
Population: 47 trials with 9,211 adults, mostly aged 35 to 60 [3]. 26 trials with 6,070 participants [4].
Sources: [3], [4]
Caveats: The two reviews come from overlapping author groups and share trials. Heterogeneity could not be explained [3].

DairyStrong

Hip fracture

Mixed results: milk was not linked to hip fracture overall in pooled cohorts, while yogurt and cheese were linked to 22% to 32% lower hip fracture.

Compared with one fewer glass of milk a day [3]; the lowest consumption category [4][5]

The numbers and sources

Effect: In women, each daily glass of milk was not linked to hip fracture (RR 0.99, 95% CI 0.96 to 1.02). In men the estimate was 0.91 (95% CI 0.81 to 1.01) [3]. In a 2020 review, the highest against the lowest intake was linked to lower hip fracture for yogurt (RR 0.78, 95% CI 0.68 to 0.90) but not milk (RR 0.86, 95% CI 0.73 to 1.02) or cheese (RR 0.85, 95% CI 0.66 to 1.08). Milk was linked to lower hip fracture in US cohorts (RR 0.75, 95% CI 0.65 to 0.87) and not in Scandinavian ones (RR 1.00, 95% CI 0.85 to 1.17) [4]. In a 2018 review of cohorts, yogurt (RR 0.75, 95% CI 0.66 to 0.86) and cheese (RR 0.68, 95% CI 0.61 to 0.77) were linked to lower hip fracture, and milk (RR 0.91, 95% CI 0.74 to 1.12) and total dairy (RR 1.02, 95% CI 0.93 to 1.12) were not [5].
Population: 195,102 women with 3,574 hip fractures and 75,149 men with 195 hip fractures in cohorts [3]. Prospective cohorts [4]. 10 cohort and 8 case-control studies [5].
Sources: [3], [4], [5]
Caveats: Observational. Yogurt eaters may follow other habits that also affect bone [4]. The US link may reflect vitamin D added to milk there, according to the review authors [4]. Calcium supplement trials are on the calcium page.

AcupunctureModerate

Chronic low back pain, compared with sham acupuncture

Mixed results: pain eased slightly more than with sham acupuncture right after treatment, a difference below the clinically important threshold, and back function did not differ.

Compared with sham acupuncture [4]

The numbers and sources

Effect: Against sham, acupuncture lowered pain by 9.22 points on a 0 to 100 scale in the immediate term (95% CI -13.82 to -4.61, 7 trials, 1,403 participants), short of the 15-point threshold the reviewers set as clinically important. Back-specific function did not differ from sham (SMD -0.16, 95% CI -0.38 to 0.06, 5 trials, 1,481 participants), and short-term quality of life did not differ by a clinically meaningful amount (SMD 0.24) [4].
Population: 33 trials with 8,270 adults with chronic nonspecific low back pain, two thirds of them in 7 German trials [4].
Sources: [4]
Caveats: Certainty was rated low to very low for these comparisons, mostly because acupuncturists could not be blinded [4].

Alkaline phosphatase (ALP)Moderate

Variation from one test to the next

Shown to vary less than other liver tests: 17% of US adults with a raised result were normal about 17 days later, and within-person variation was 5% to 7%.

The numbers and sources

Compared with: A repeat measurement in the same person
Effect: Among 1,864 US adults measured twice a mean of 17.5 days apart, 17% of those with an initially elevated alkaline phosphatase had a normal result the second time, against 31% for ALT and 38% for bilirubin. More than 95% of initially normal results stayed normal. The within-person coefficient of variation was 6.7% [12]. In 91 European adults sampled weekly for 10 weeks under a tightly controlled protocol, within-person biological variation was 5.3% and variation between people was 24.9% [13].
Population: 1,864 adults in the third US National Health and Nutrition Examination Survey [12]. 91 adults aged 21 to 69 at 6 European laboratories [13].
Sources: [12], [13]
Caveats: The US study had only 2 measurements per person and lacked a full liver disease history [12]. Neither study states a threshold for agreement. One survey manual notes that hemoglobin from a hemolyzed sample lowers the result [4].

AllopurinolModerate

Gout flares

Mixed results: no fewer gout attacks than placebo over 30 days in one small trial, and flares became less frequent with continued treatment in longer trials.

The numbers and sources

Effect: Over 30 days, gout attacks occurred in 2 of 26 people on allopurinol 300 mg and 3 of 25 on placebo (RR 0.64, 95% CI 0.12 to 3.52) [1]. Over weeks 9 to 52 of a year-long trial, 64% on allopurinol had a flare, and the incidence fell with continued treatment [2]. In the final 24 weeks of a 72-week trial, 36.5% on allopurinol had one or more flares against 43.5% on febuxostat, meeting the test for noninferiority [4]. The label states that gout attacks usually become shorter and less severe after several months of treatment [5].
Population: 51 people in the placebo-controlled trial [1]; 760 people with gout [2]; 940 people with gout [4].
Sources: [1], [2], [4], [5]
Caveats: The longer trials had no placebo group, so the fall in flares over time has no untreated comparison [2][4].

CalciumModerate

Fractures

Mixed results: recent reviews found calcium alone did not lower fractures and calcium with vitamin D lowered them only slightly. Older reviews found about 11% to 12% fewer.

The numbers and sources

Effect: Recent meta-analyses found little or no effect, and an older one found fewer fractures. A 2026 meta-analysis of 69 trials (153,902 adults) found calcium alone did not significantly lower the risk of any fracture (RR 0.91, 95% CI 0.81 to 1.01, 11 trials, 9,067 people, moderate certainty). Calcium plus vitamin D lowered it slightly (RR 0.91, 95% CI 0.84 to 0.99, 15 trials, 51,126 people, high certainty), below the authors' threshold for a meaningful absolute effect. Other fracture and fall outcomes showed little to no effect [1]. In postmenopausal women with osteoporosis, calcium plus vitamin D did not lower fracture risk (RR 0.98, 95% CI 0.89 to 1.07, 11 RCTs, 43,869 women) [2]. A 2007 meta-analysis of 17 trials (52,625 people) found 12% fewer fractures with calcium or calcium plus vitamin D (RR 0.88, 95% CI 0.83 to 0.95), 24% fewer where adherence was high [3]. A 2015 review found 11% fewer total fractures (RR 0.89, 95% CI 0.81 to 0.96) and no change in hip fractures (RR 0.95, 95% CI 0.76 to 1.18). In the 4 trials at lowest risk of bias (44,505 people), there was no effect at any site, and the authors found signs of publication bias toward calcium [4]. In community-dwelling adults over 50, calcium did not lower hip fracture (RR 1.53, 95% CI 0.97 to 2.42) or other fractures [5].
Population: Adults not on osteoporosis drugs, 87% living in the community [1]. Postmenopausal women with osteoporosis [2]. Adults aged 50 or older in 17 trials [3], 26 trials [4] and 33 trials with 51,145 people [5].
Sources: [1], [2], [3], [4], [5]
Caveats: The 2015 review found only one trial with a clear fracture reduction, in frail women in residential care with low calcium intake and low vitamin D [4]. Evidence for people in residential care or at high fracture risk was limited in the 2026 review [1].

Chiropractic and spinal manipulationModerate

Chronic low back pain, compared with sham manipulation

Mixed results: one small meta-analysis found manipulation eased low back pain more than sham, and two larger reviews rated the sham evidence too weak to settle.

Compared with sham manipulation [3][4][5]; inert interventions [4]

The numbers and sources

Effect: In 4 trials pooled against sham, symptoms improved more with manipulation (SMD -0.36, 95% CI -0.59 to -0.12) [5]. A review of 47 trials rated the evidence against sham low to very low in quality and called its effects uncertain [3]. An earlier Cochrane review found very low-quality evidence that manipulation was not significantly better than inert interventions or sham for short-term pain or function [4].
Population: 9 trials, 4 pooled, in nonspecific low back pain [5]. 9,211 adults [3]. 6,070 participants [4].
Sources: [5], [3], [4]
Caveats: The practitioner cannot be blinded, several sham trials were small, and some had high dropout [5].

CuppingModerate

Chronic pain, compared with other active treatments

Mixed results: one meta-analysis found no pain difference from other active treatments, while reviews of neck and low back pain trials found cupping ahead.

Compared with other active treatments [2][7]; medication [8]; usual care [8]

The numbers and sources

Effect: Against other active treatments, chronic pain did not differ significantly (SMD -0.24, 95% CI -0.57 to 0.09), while disability was lower (SMD -0.52) [2]. In neck pain trials, pain fell more than with active controls (p = 0.0009) [7]. In low back pain trials, pain fell more than with medication (d 1.8, 8 trials) and usual care (d 1.07, 5 trials) at 2 to 8 weeks, with no difference at 1 month or 3 to 6 months [8].
Population: 18 trials [2]. 18 neck pain trials [7]. 921 participants in 11 low back pain trials [8].
Sources: [2], [7], [8]
Caveats: Heterogeneity was high across the low back pain subgroups [8].

Dry needlingModerate

Musculoskeletal pain, compared with sham needling

Mixed results: in adequately blinded trials, pain did not differ from sham needling. Less well blinded sham trials favored dry needling, as did two small low back pain trials.

Compared with sham dry needling, penetrating or non-penetrating [5]; sham needling [6]

The numbers and sources

Effect: Across 24 sham-controlled trials, the 3 adequately blinded comparisons found no difference in short-term pain (SMD -0.40, 95% CI -0.96 to 0.15, 122 participants), and the 4 adequately blinded comparisons at 1 to 6 months found none in long-term pain (SMD -0.30, 95% CI -0.62 to 0.02, 202 participants). Inadequately blinded comparisons favored dry needling in the short term (SMD -0.71, 95% CI -1.05 to -0.38, 504 participants) and the long term (SMD -1.14, 95% CI -1.64 to -0.65, 284 participants). The one adequately blinded comparison measured right after treatment, in 31 people, favored dry needling (SMD -0.76) [5]. In low back pain, 2 sham-controlled trials favored dry needling for pain after treatment (SMD -2.74, 95% CI -3.77 to -1.71) and at follow-up (SMD -1.05, 95% CI -1.70 to -0.40), and disability at follow-up did not differ [6].
Population: 24 trials comparing dry needling with a sham that simulated it, searched to February 2016 [5]. 2 sham-controlled trials within 16 low back pain trials [6].
Sources: [5], [6]
Caveats: 19 of the 24 trials were at high or unclear risk of bias, and no trial tried to blind the therapist. The reviewers say the subgroups were small and the results are uncertain [5]. Most low back pain trials were at high or unclear risk of bias [6].

Dry needlingModerate

Low back pain, compared with acupuncture

Mixed results: low back pain eased more than with acupuncture right after treatment in 6 trials, and pain and disability did not differ at follow-up.

Compared with acupuncture, including wrist-ankle acupuncture in one trial [6]

The numbers and sources

Effect: Across 6 trials with 467 participants, pain intensity after treatment was lower with dry needling than with acupuncture (SMD -0.96, 95% CI -1.80 to -0.12). At follow-up, pain and disability did not differ. In one trial of 60 people with acute low back pain, one 5-minute session lowered pain more than wrist-ankle acupuncture [6].
Population: 467 people with low back pain in 6 trials [6].
Sources: [6]
Caveats: Most trials were at high or unclear risk of bias, and the reviewers say the evidence is not robust [6].

EstradiolModerateStudied in women

Fractures, stroke and breast cancer in trials of estrogen therapy after menopause

Mixed results: in one large trial estrogen alone cut fractures by 27% and raised strokes by 33% in postmenopausal women, with little or no difference in breast cancer.

The numbers and sources

Effect: A Cochrane review drew its main results from the Women's Health Initiative. Among 10,739 women who had had a hysterectomy, oral conjugated equine estrogen for an average of 7 years, against placebo, gave relative risks of 0.73 for all clinical fractures (95% CI 0.65 to 0.80), 1.33 for stroke (1.06 to 1.67), 1.78 for gallbladder disease requiring surgery (1.42 to 2.24), 0.94 for coronary events (0.78 to 1.13), 1.32 for venous clots (1.00 to 1.74) and 0.79 for breast cancer (0.61 to 1.01). Among 16,608 women with a uterus, estrogen with a progestin for 5.6 years gave 0.78 for fractures (0.71 to 0.86), 1.39 for stroke (1.09 to 2.09), 2.03 for venous clots (1.55 to 6.64) and 1.27 for breast cancer (1.03 to 1.56) [10].
Population: 24 randomized trials with 45,660 women, most of them postmenopausal American women with a mean age over 60 [10].
Sources: [10], [7]
Caveats: A marker that tracks an outcome is not always one whose correction changes it. Higher estradiol went with more breast cancer in cohorts [7], and estrogen given alone did not raise it in this trial. The trials set a dose and did not target a blood level. The review's authors caution that the results rest on one study of oral therapy, and that only about 30% of the women were aged 50 to 59 [10].

29 more

Fish oilModerate

Rheumatoid arthritis

Mixed results: one review found better disease activity markers in rheumatoid arthritis, while others found small or no effects on pain and swollen joints.

The numbers and sources

Effect: Across 20 RCTs (1,252 patients), omega-3 improved eight disease activity markers. Of five inflammatory markers, only leukotriene B4 fell (SMD -0.440, 95% CI -0.676 to -0.205) [11]. Across 23 placebo-controlled trials, pooled effects on pain (SMD -0.16, 95% CI -0.40 to 0.09), tender joints (SMD -0.20, 95% CI -0.46 to 0.05) and swollen joints (SMD -0.10, 95% CI -0.28 to 0.07) were small, with intervals that include no effect [12]. Across 18 RCTs (1,018 patients), tender joint count fell (SMD -0.59, 95% CI -0.79 to -0.39), while ESR, CRP and DAS28 did not change significantly [13].
Population: Adults with rheumatoid arthritis in RCTs [11][12][13].
Sources: [11], [12], [13]
Caveats: The certainty of the evidence was mostly very low or low [12]. The authors of [12] note that earlier findings of lower NSAID use may reflect inadequate blinding.

Fish oilModerate

Chronic pain and knee osteoarthritis

Mixed results: omega-3 lowered chronic pain but not osteoarthritis pain in one review, while krill oil reviews found better knee osteoarthritis pain scores.

The numbers and sources

Effect: Across 41 RCTs (3,759 participants), omega-3 lowered chronic pain intensity (SMD -0.55, 95% CI -0.76 to -0.34, I² 87%), from -0.27 at 1 month to -0.83 at 6 months. Effects were significant in rheumatoid arthritis, migraine and mixed chronic pain, and not in osteoarthritis or breast pain [27]. In knee osteoarthritis, krill oil improved WOMAC pain (SMD -0.60, 95% CI -0.99 to -0.21), stiffness and function across 5 RCTs (730 participants), with no change on a visual analog pain scale [28]. A 2026 review of 6 RCTs (971 participants) also found better WOMAC pain and function at week 4 [29].
Population: Adults with chronic pain conditions [27]; adults with knee osteoarthritis [28][29].
Sources: [27], [28], [29]
Caveats: Heterogeneity was high in [27]. The reviews disagree on osteoarthritis, though the krill reviews tested one product type. Lower doses (1.35 g a day or less) did as well as higher doses [27].

Glucosamine and chondroitinModerate

Glucosamine product type

Mixed results: trials of the Rotta preparation showed larger pain relief, while other preparations and well-concealed trials found no benefit.

The numbers and sources

Effect: Across 20 RCTs, glucosamine improved pain (SMD -0.61, 95% CI -0.95 to -0.28), but analyses limited to 8 trials with adequate allocation concealment found no benefit. Trials of the Rotta preparation showed pain SMD -1.31, while non-Rotta preparations showed SMD -0.15 (95% CI -0.35 to 0.05) [5].
Population: 2,570 patients in 20 RCTs [5].
Sources: [5]
Caveats: WOMAC pain, function and stiffness did not differ from placebo in this review. The 2005 search predates the large independent trials [2][3].

Glucosamine and chondroitinModerate

Joint space narrowing

Mixed results: one review found minute differences in joint space width, while others found slightly less narrowing with chondroitin or Rotta glucosamine.

The numbers and sources

Effect: Differences in joint space width were minute, with credible intervals overlapping zero for all three preparations [1]. In 2 trials, loss of minimum joint space width was 4.7% less with chondroitin (95% CI 1.6% to 7.8%) [4]. Two Rotta glucosamine trials found slower radiological progression over 3 years (SMD 0.24) [5].
Population: Adults with knee osteoarthritis followed up to 3 years [1][4][5].
Sources: [1], [4], [5]
Caveats: Reviews disagree. Joint space width is a structural measure and was not tied to symptoms in these abstracts.

Osteopathy and osteopathic manipulative treatmentModerate

Low back pain, compared with sham treatment

Mixed results: a 2024 review of sham-controlled trials found no difference in back or neck pain, while the largest trial found more people improved with treatment than with sham.

Compared with sham or placebo interventions [2]; sham osteopathic treatment [3]

The numbers and sources

Effect: Pooled against sham or placebo, osteopathic treatment did not change pain intensity (SMD -0.15, 95% CI -0.38 to 0.08, 7 trials, 1,173 patients) or disability (SMD -0.09, 95% CI -0.25 to 0.08, 6 trials, 1,153 patients) in neck or low back pain, and most of the 9 trials found no difference [2]. In the largest trial, people given osteopathic treatment were more likely than those given sham to have a 30% or greater drop in chronic low back pain at 12 weeks (response ratio 1.38, 95% CI 1.16 to 1.64) and a 50% or greater drop (1.41, 95% CI 1.13 to 1.76), a medium effect by Cochrane Back Review Group criteria. Back function, general health and work disability did not differ from sham [3].
Population: 9 trials of adults with neck or low back pain [2]. 455 adults with chronic nonspecific low back pain [3].
Sources: [2], [3]
Caveats: The 2024 review pooled neck and back pain and rated the certainty moderate to very low [2]. The practitioner cannot be blinded to giving real or sham treatment.

Osteopathy and osteopathic manipulative treatmentModerateStudied in women

Low back pain in pregnancy and after childbirth

Mixed results: a review found less pain in pregnant and postpartum women, while the largest pregnancy trial found treatment no different from placebo ultrasound, with both better than usual care.

Compared with usual care only, and usual care plus placebo ultrasound [7]; control groups the source abstract does not describe [8]

The numbers and sources

Effect: In pregnancy, pain fell 16.65 points more (on a 100-point scale) and function improved (SMD -0.50), at moderate quality. After childbirth, pain fell 38.00 points more and function improved (SMD -2.12), at low quality [8]. In 400 women in the third trimester, 7 treatments over 9 weeks eased pain and slowed loss of back function more than usual care alone (P < .001), and results with osteopathic treatment were similar to placebo ultrasound [7].
Population: 5 trials in pregnancy and 3 after childbirth [8]. 400 women in their third trimester [7].
Sources: [8], [7]
Caveats: The trial authors suggest the placebo ultrasound may have been more active than intended [7].

Osteopathy and osteopathic manipulative treatmentModerate

Headache, irritable bowel syndrome and other conditions

Mixed results: an overview of 27 reviews found possible relief in migraine and irritable bowel syndrome, and no relief in acute low back pain, fibromyalgia or other headaches.

Compared with no treatment, sham treatment or other non-osteopathic treatment [16]

The numbers and sources

Effect: Osteopathic treatment may relieve symptoms in adults with migraine and irritable bowel syndrome. It did not appear to relieve acute nonspecific low back pain, fibromyalgia, headaches of various causes, lower urinary tract symptoms in women, or cerebral palsy in children. Evidence for vertigo, COPD, high blood pressure and gynecological complaints was insufficient [16]. An earlier overview rated the evidence for children's conditions, headache and irritable bowel syndrome limited and inconclusive [17].
Population: 27 systematic reviews covering nine indications [16]. 9 reviews with 55 trials and 3,740 participants [17].
Sources: [16], [17]
Caveats: Included reviews were rated low or critically low in quality [17], and the trials were heterogeneous and partly of poor quality [16].

Ozone therapyModerate

Low back pain, compared with usual care and other treatments

Mixed results: pooled trials found a small function gain at one month against usual care, no lasting pain difference, and no difference in excellent or poor outcomes.

Compared with usual care and other treatments the review abstract does not list [4]

The numbers and sources

Effect: Across 9 studies, disability on the Oswestry index favored ozone overall (SMD -0.28, 95% CI -0.51 to -0.06), at one month only, with no difference at 2 weeks or 6 months. Pain favored ozone overall by a small amount (SMD -0.12, 95% CI -0.24 to -0.01), and subgroup and sensitivity analyses found no difference. Excellent outcomes (OR 0.95, 95% CI 0.54 to 1.67) and poor outcomes (OR 1.25, 95% CI 0.66 to 2.37) did not differ [4].
Population: 1,711 participants with low back pain in 9 studies [4].
Sources: [4]
Caveats: The reviewers describe the benefit over usual care as limited and the results as inconsistent [4].

Ozone therapyModerate

Joint and muscle pain, compared with corticosteroid injections

Mixed results: across musculoskeletal conditions, steroid injections eased pain more in the short term, while in knee osteoarthritis alone ozone eased pain more.

Compared with corticosteroid injections [8][9]

The numbers and sources

Effect: Across 11 studies of musculoskeletal pain of several kinds, short-term pain fell more with corticosteroid injections (d 0.31, 95% CI 0.01 to 0.60), and medium-term pain did not differ (d -0.17, 95% CI -0.42 to 0.07) [8]. In 7 knee osteoarthritis trials, ozone injections reduced pain more than corticosteroid injections in the short and medium term and improved function in the medium term [9].
Population: 534 people with musculoskeletal pain in 11 studies [8]. 409 people with knee osteoarthritis in 7 trials [9].
Sources: [8], [9]
Caveats: Both reviews describe the included trials as limited in quality [8][9]. The knee review abstract gives no effect sizes [9].

Physical therapy and exercise therapyModerate

Chronic low back pain, compared with other treatments

Mixed results: exercise eased pain about 9 points on a 100-point scale more than other conservative treatments, below the clinical threshold, and did not differ from manual therapy.

Compared with other conservative treatments [2]; education alone [2]; non-exercise physical therapy [2]; manual therapy [2]

The numbers and sources

Effect: Against all other conservative treatments, pain fell 9.1 points more (95% CI -12.6 to -5.6) and functional limitation 4.1 points more, neither meeting the reviewers' threshold for clinical importance. By subgroup, pain fell more than with education alone (-12.2) and non-exercise physical therapy (-10.4), and did not differ from manual therapy (1.0, 95% CI -3.1 to 5.1) [2].
Population: 249 trials of adults with chronic low back pain [2].
Sources: [2]
Caveats: Low-certainty evidence for these comparisons [2].

Physical therapy and exercise therapyModerate

Compared with surgery

Mixed results: trials found outcomes similar to surgery for degenerative meniscus tears, spinal stenosis and ACL tears, and smaller gains than knee replacement.

Compared with arthroscopic partial meniscectomy [9]; surgical decompression [10]; early ACL reconstruction [11]; total knee replacement followed by nonsurgical treatment [12]

The numbers and sources

Effect: For degenerative meniscus tears, knee scores at 2 years differed by 0.9 points between 12 weeks of supervised exercise and surgery (95% CI -4.3 to 6.1), and 19% of the exercise group crossed over to surgery [9]. For lumbar spinal stenosis, physical function improved by 19.2 points with physical therapy and 22.4 with surgery, with no difference at 2 years, and 57% crossed over from therapy to surgery [10]. For acute ACL tears, knee scores rose 39.4 points with rehabilitation plus optional later reconstruction and 39.2 with rehabilitation plus early reconstruction, and 36 of 59 never had surgery [11]. For knee osteoarthritis, knee replacement plus nonsurgical treatment improved scores by 32.5 points against 16.0 with nonsurgical treatment alone, and had more serious adverse events (24 against 6) [12].
Population: 140 adults aged 35 to 60 with degenerative meniscal tears [9]. 169 surgical candidates aged 50 or older with lumbar spinal stenosis [10]. 121 young, active adults with acute ACL tears [11]. 100 patients eligible for knee replacement [12].
Sources: [9], [10], [11], [12]
Caveats: Each condition rests on one trial, and crossover to surgery was common [9][10]. The spinal stenosis trial had no untreated group [10].

ProbioticsModerate

Rheumatoid arthritis

Mixed results: reviews agree probiotics lower CRP or IL-6 in rheumatoid arthritis, but disagree on disease activity.

The numbers and sources

Effect: Across 9 studies (361 patients), probiotics lowered IL-6 (SMD -0.708) and did not change disease activity scores (MD 0.023, 95% CI -0.584 to 0.631) [8]. Across 4 RCTs (153 participants), DAS fell (SMD -0.58, 95% CI -0.97 to -0.19), CRP fell borderline (SMD -0.32, 95% CI -0.65 to 0.00), and ESR and cytokines did not change [9]. A review of 8 RA trials (344 patients) found CRP 3.04 mg/L lower (95% CI -4.47 to -1.62). With high risk-of-bias trials excluded, DAS28 did not differ from placebo (SMD -0.54, 95% CI -1.94 to 0.85) [10].
Population: Adults with rheumatoid arthritis, mostly women, over 8 weeks to 1 year [8][9][10].
Sources: [8], [9], [10]
Caveats: The reviews agree on lower CRP or IL-6 and disagree on disease activity. Strains, doses and durations varied [10].

PRP injectionsModerate

Knee osteoarthritis, compared with saline placebo

Mixed results: pooled placebo-controlled trials found less knee pain up to 6 months, while the largest trial, of 288 people, found no difference from saline at 12 months.

Compared with saline placebo injections [2][3][4]

The numbers and sources

Effect: In 288 adults aged 50 or older, knee pain changed by -2.1 points with PRP and -1.8 with saline at 12 months (difference -0.4, 95% CI -0.9 to 0.2, on a 0 to 10 scale with a clinically important difference of 1.8). Cartilage volume did not differ, and 29 of 31 secondary outcomes did not differ [2]. Across 11 trials, pain was lower than placebo at 3 months (MD -1.1, 95% CI -1.8 to -0.4) and 6 months (MD -2.0, 95% CI -2.8 to -1.2), and not at 12 months. WOMAC scores improved at 3 months only [3]. A review of 20 placebo-controlled studies found better function and pain at 6 and 12 months, at very low certainty [4].
Population: 288 adults with mild to moderate medial knee osteoarthritis in Australia [2]. 1,616 patients in 11 trials, mean age 56.9 [3]. 1,774 PRP and 1,518 placebo participants in 20 studies [4].
Sources: [2], [3], [4]
Caveats: Heterogeneity was high in the pooled pain results [3], and certainty very low [4]. The largest trial, blinded for participants, injectors and assessors, found no difference [2].

PRP injectionsModerate

Rotator cuff tendinopathy, injection

Mixed results: a pooled review reported short-term pain relief against placebo and corticosteroids, while a placebo-controlled trial found no difference from saline at one year.

Compared with placebo or corticosteroid injections [10]; saline injection, both groups exercising [11]

The numbers and sources

Effect: Across 30 randomized trials, PRP reduced pain and improved function in the short term against placebo and corticosteroids, with no greater benefit beyond 12 months [10]. In 40 patients, WORC, SPADI and pain scores did not differ from saline at any time point through one year (P = .174, .314 and .904 at one year), and both groups improved [11].
Population: About 2,500 participants in 30 trials [10]. 40 adults aged 18 to 70 with chronic rotator cuff tendinopathy [11].
Sources: [10], [11]
Caveats: The review abstract gives no effect sizes and does not separate placebo from corticosteroid comparisons [10].

Red light therapyModerate

Low back pain

Mixed results: one meta-analysis found pain reduction with higher-dose laser in chronic low back pain, and a later one found the effect against sham clinically unimportant.

The numbers and sources

Effect: Across 15 sham-controlled studies (1,039 participants), pain fell by up to 1.40 cm on a 10 cm scale (95% CI -1.91 to -0.88) in trials using at least 3 J per point and in people with pain for under 30 months [15]. Across 12 RCTs (1,046 participants), the effect on pain and disability compared with sham was clinically unimportant in acute, subacute and chronic low back pain, and did not differ meaningfully from exercise [16].
Population: Adults with chronic non-specific low back pain [15]. People with acute, subacute or chronic non-specific low back pain [16].
Sources: [15], [16]
Caveats: The earlier review's benefit appeared only in dose and duration subgroups [15]. The later review concluded current evidence does not support its use for this condition [16].

Strength trainingModerate

Falls in older adults

Mixed results: programs combining balance and strength exercises probably cut falls by about a third, while the effect of strength training alone was uncertain.

Compared with control interventions not thought to reduce falls [16]

The numbers and sources

Effect: Programs with several exercise types, most often balance and functional exercises plus resistance exercises, probably reduced the rate of falls by 34% (RaR 0.66, 95% CI 0.50 to 0.88) and the number of people falling by 22% (RR 0.78). Balance and functional exercises reduced the rate of falls by 24%. The review was uncertain of the effect of programs that were mainly resistance training [16].
Population: 23,407 people 60 and older living in the community, average age 76, in 108 trials [16].
Sources: [16]
Caveats: Evidence for mixed programs was rated moderate in certainty, and for resistance training alone too uncertain to judge [16].

TENSModerate

Chronic low back pain

Mixed results: placebo-controlled trials conflicted for chronic low back pain. A 2026 review found an effect only when stimulation was strong enough to feel clearly.

Compared with placebo TENS [10]; control or placebo groups [11]

The numbers and sources

Effect: Four high-quality trials gave conflicting results on pain against placebo, and two (410 patients) found no improvement in back function [10]. Across 29 studies, TENS had no overall effect on pain (d 0.16, 64 comparisons). Against control or placebo, trials using a clearly felt, adjusted intensity showed a larger effect (d 0.97, 95% CI 0.65 to 1.30) than trials using too little (d 0.30, 95% CI -0.11 to 0.72) [11].
Population: 585 patients in 4 trials [10]. 29 studies [11].
Sources: [10], [11]
Caveats: The later review rated certainty low because of high risk of bias and heterogeneity [11].

Volunteering and purposeModerate

Physical function in older volunteers

Mixed results: one meta-analysis linked volunteering after 65 to 17% less functional disability, while an earlier review found no favorable effect on physical health.

Compared with non-volunteers [3]; non-volunteers [2]; people who volunteered 0 hours a year [4]

The numbers and sources

Effect: Volunteers had less new functional disability (HR 0.83, 95% CI 0.72 to 0.97) and kept functional competence more often (OR 0.81, 95% CI 0.70 to 0.94). Decline in daily living tasks did not differ significantly (OR 0.73, 95% CI 0.53 to 1.01) [3]. Cohort studies did not show favorable effects on physical health [2]. Volunteering 100 hours or more a year was linked to fewer physical functioning limitations and more physical activity [4].
Population: Adults 65 and older in 3 disability and 3 functional competence studies [3]. 17 cohort studies [2]. 12,998 US adults over 50 [4].
Sources: [3], [2], [4]
Caveats: Observational. People with better function are more able to start and keep volunteering, which is reverse causation.

BisphosphonatesLimitedStudied in women

Continuing alendronate past 5 years

One trial found women who stopped alendronate after 5 years had the same non-vertebral fracture rate as those who continued, but more clinical spine fractures, 5.3% against 2.4%.

Compared with switching to placebo after 5 years [14]

The numbers and sources

Effect: Over 5 more years, non-vertebral fractures were 19% in women who continued and 18.9% in those who switched to placebo (RR 1.00, 95% CI 0.76 to 1.32). Clinical spine fractures were 2.4% against 5.3% (RR 0.45, 95% CI 0.24 to 0.85), and spine fractures on x-ray did not differ. Hip bone density fell 2.4% more after stopping [14].
Population: 1,099 postmenopausal women who had taken alendronate for a mean 5 years [14].
Sources: [14]
Caveats: Fracture was an exploratory outcome. One trial.

BromelainLimited

Joint pain and osteoarthritis

Limited evidence: one trial found no change in knee osteoarthritis scores against placebo, and pooled studies across conditions found slightly lower pain.

The numbers and sources

Effect: In 47 adults with moderate to severe knee osteoarthritis, 800 mg a day for 12 weeks did not change the total WOMAC score against placebo (coefficient 11.16, 95% CI -8.86 to 31.18, p 0.27) [6]. In 40 adults with mild to moderate knee osteoarthritis, 500 mg a day did not differ from diclofenac 100 mg a day at 4 weeks [7]. Across 9 studies of oral bromelain in various conditions, pain scores were slightly lower than in controls (MD -0.27, 95% CI -0.45 to -0.08, I² 29%) [8]. An earlier review found that all trials before 2004 were uncontrolled or comparative [9].
Population: Adults with knee osteoarthritis [6][7]. Mixed conditions [8].
Sources: [6], [7], [8], [9]
Caveats: Only 31 of 47 finished the placebo trial [6]. The diclofenac trial was single-blind with no placebo [7]. The pooled pain estimate mixes conditions [8].

Calcium levelLimited

Published severity grades use corrected calcium

Two endocrine societies grade high calcium on albumin-adjusted calcium, so their cut points of 12 and 14 mg/dL are not ranges for the total calcium on a standard panel.

The numbers and sources

Compared with: Albumin-adjusted calcium, the quantity these guidelines grade
Effect: The Endocrine Society guideline on hypercalcemia of malignancy classifies albumin-adjusted serum calcium below 12 mg/dL (3 mmol/L) as mild, 12 to 14 mg/dL (3 to 3.5 mmol/L) as moderate and above 14 mg/dL (3.5 mmol/L) as severe [11]. UK Society for Endocrinology guidance states that calcium measurements are to be adjusted for serum albumin, and describes levels below 3.0 mmol/L as often without symptoms, 3.0 to 3.5 mmol/L as tolerated when the rise has been slow, and above 3.5 mmol/L as carrying a risk of dysrhythmia and coma [12]. Its companion guidance on low calcium states that symptoms typically develop when adjusted calcium falls below about 1.9 mmol/L, and calls severe hypocalcemia a medical emergency [13].
Population: Adults with hypercalcemia of malignancy [11]. Adult patients seen by non-specialists in the UK [12][13].
Sources: [11], [12], [13]
Caveats: These grades are for a calculated quantity, so this page does not place a typed total calcium against them. The Endocrine Society introduces the same cut points under the plain term serum calcium and specifies albumin-adjusted calcium in its treatment figure [11]. The UK guidance notes that the threshold for symptoms varies greatly and depends on how fast calcium falls [13].

CollagenLimited

Bone density

Limited evidence: one review found higher bone density, partly with added calcium and vitamin D, while another found the bone studies too limited to conclude.

The numbers and sources

Effect: A 2025 meta-analysis found collagen peptides raised bone mineral density at the femoral neck and spine (I² 80.1%) and improved bone turnover markers (SMD 0.40 to 0.58) [10]. A systematic review of 36 RCTs found limitations in the bone studies that prevent definitive conclusions [4]. In 39 older adults after total hip replacement, 15 g a day of collagen for 14 days did not change bone turnover markers [13].
Population: People in RCTs, some given calcium and vitamin D with collagen [10]; RCTs of type I collagen alone [4]; older adults after hip surgery [13].
Sources: [10], [4], [13]
Caveats: Sources disagree. Heterogeneity for bone density was high, and the meta-analysis pooled collagen alone and with calcium and vitamin D [10]. The abstract does not report fracture results [10].

Infrared saunaLimited

Rheumatoid arthritis and ankylosing spondylitis

Mixed results: pain and stiffness eased during infrared sessions in 34 patients, while changes across the 4-week course did not reach significance.

The numbers and sources

Effect: Pain and stiffness decreased significantly during a session (p < 0.05 in rheumatoid arthritis and p < 0.001 in ankylosing spondylitis). Over 4 weeks and 8 sessions, improvements in pain, stiffness and fatigue did not reach statistical significance. Disease activity scores did not change [27].
Population: 17 people with rheumatoid arthritis and 17 with ankylosing spondylitis, no control group [27].
Sources: [27]
Caveats: A pilot study with no control group.

Massage therapyLimited

Neck pain

Limited evidence: very low-quality trials suggested some massage techniques improved function and tenderness more than control or placebo, and reviewers called the effect uncertain.

Compared with control or placebo treatment [6]; education [6]

The numbers and sources

Effect: Very low-quality evidence suggested traditional Chinese massage and strain-counterstrain techniques may improve function and tenderness more than control or placebo, and massage may ease pain bothersomeness more than education in the short term. Most trials measured outcomes immediately after treatment [6].
Population: 15 trials of adults with mechanical neck disorders [6].
Sources: [6]
Caveats: All trials were rated low or very low in quality, and only 4 of 15 described the massage adequately. The reviewers made no recommendation [6].

Osteopathy and osteopathic manipulative treatmentLimited

Neck pain

Limited evidence: four small trials pooled against no treatment or sham found less neck pain, at very low quality, and a 2024 sham-only review found no difference.

Compared with no intervention or sham treatment, pooled together [9]; sham or placebo interventions [2]

The numbers and sources

Effect: Against no intervention or sham, osteopathic treatment lowered pain (effect size -1.57, 95% CI -2.50 to -0.65) and improved function (-1.71), with very low quality evidence for every outcome [9]. The 2024 review, which pooled neck and low back pain against sham or placebo only, found no difference in pain or disability [2].
Population: 5 trials, 4 pooled, of adults with nonspecific neck pain [9]. 9 trials in neck or low back pain [2].
Sources: [9], [2]
Caveats: None of the neck pain trials was judged at low risk of bias in every domain [9].

PotassiumLimited

Bone

Limited evidence: alkaline potassium salts lowered a bone breakdown marker, with bone density unchanged in a review and up 1.7% in one trial. No fracture data.

The numbers and sources

Effect: Sources disagree. A meta-analysis of 14 studies found potassium bicarbonate and citrate lowered urinary calcium and the bone breakdown marker NTX, with no effect on bone density [11]. In one 2-year trial, 60 mEq of potassium citrate a day raised lumbar spine density 1.7% over placebo (95% CI 1.0 to 2.3) [12]. In 5 RCTs of postmenopausal women, potassium citrate lowered urinary NTX at 1 month (MD -4.05 nmol BCE/mmol creatinine, 95% CI -8.05 to -0.05). CTX, P1NP, urinary calcium and parathyroid hormone did not differ from placebo [13].
Population: 14 studies of alkaline potassium salts [11]; 201 healthy adults over 65 without osteoporosis, all given calcium and vitamin D [12]; 401 postmenopausal women [13].
Sources: [11], [12], [13]
Caveats: No fracture outcomes. Potassium chloride, the form in salt substitutes, lowered urinary calcium less than the alkaline salts [11].

Reiki and therapeutic touchLimited

Pain, compared with sham Reiki

Mixed results: a fibromyalgia trial found Reiki no different from sham given by actors, and a knee pain trial found more people improved with Reiki than with sham.

Compared with sham Reiki given by actors, with or without touch [7]; sham Reiki and a waiting list [8]

The numbers and sources

Effect: In 100 adults with fibromyalgia treated twice weekly for 8 weeks, pain and every secondary outcome were nearly identical whether a Reiki master or an actor gave the treatment, with touch or at a distance [7]. In 164 adults with knee osteoarthritis, symptoms improved more with Reiki than with a waiting list (P = .02), and a reduction of more than 30% was reported by 55% given Reiki, 20% given sham Reiki and 13% on the waiting list [8].
Population: 100 adults with fibromyalgia in Seattle [7]. 164 adults with knee osteoarthritis [8].
Sources: [7], [8]
Caveats: The knee trial abstract does not report a direct test of Reiki against sham [8].

Vitamin K2Limited

Fractures

Limited evidence: older Japanese trials found fewer fractures, but a 2019 update found no difference in spine fractures and no significant effect in low-bias trials.

The numbers and sources

Effect: Sources disagree. A 2006 meta-analysis of 7 Japanese MK-4 trials found lower odds of vertebral fractures (OR 0.40, 95% CI 0.25 to 0.65), hip fractures (OR 0.23, 95% CI 0.12 to 0.47) and all nonvertebral fractures (OR 0.19, 95% CI 0.11 to 0.35) [1]. A 2019 update found no difference in vertebral fractures (OR 0.96, 95% CI 0.83 to 1.11). Any clinical fracture was lower (OR 0.72, 95% CI 0.55 to 0.95), but in low-bias trials the difference was not significant (OR 0.76, 95% CI 0.58 to 1.01) [2].
Population: Japanese patients in all 7 fracture trials [1]; postmenopausal or osteoporotic adults in 36 studies [2].
Sources: [1], [2]
Caveats: The 2019 authors write that the integrity of some earlier evidence had been questioned since the 2006 review [2]. Both reviews pooled vitamin K1 and K2 trials for bone density.

Vitamin K2Limited

Bone mineral density

Limited evidence: spine density gains vanished after removing high-bias trials. One 3-year trial found slower bone loss in postmenopausal women.

The numbers and sources

Effect: Pooled lumbar spine density was 1.63% higher at 2 years (95% CI 0.10 to 3.16), but removing high-bias trials made the differences smaller and not significant. The 2019 authors conclude there is no evidence of an effect on bone density in postmenopausal or osteoporotic patients [2]. In one 3-year trial, 180 µg of MK-7 a day slowed the age-related loss of density at the lumbar spine and femoral neck, not the total hip [3].
Population: Postmenopausal or osteoporotic adults [2]; 244 healthy postmenopausal women in the Netherlands [3].
Sources: [2], [3]
Caveats: The MK-7 trial abstract gives no effect sizes [3]. Too few trials in men or other groups for conclusions [2].

Not shown to help33

AcetaminophenStrong

Acute low back pain

Not shown to lower pain, disability or time to recovery in adults with acute low back pain at up to 4 g a day.

The numbers and sources

Effect: In a trial of 1,652 people, median time to recovery was 17 days on regular paracetamol, 17 days on as-needed paracetamol and 16 days on placebo (regular against placebo HR 0.99, 95% CI 0.87 to 1.14). Adverse events were reported by 18.5%, 18.7% and 18.5% [10]. Pooled, paracetamol did not lower pain intensity (mean difference -0.5 on a 0 to 100 scale, 95% CI -2.9 to 1.9) or disability (0.4, 95% CI -1.7 to 2.5) in the short term [9]. A Cochrane review found no difference from placebo at 1, 2, 4 and 12 weeks in pain, function, quality of life or sleep quality at 4 g a day [11].
Population: 1,652 adults with acute low back pain, 550 on regular dosing, 549 on as-needed dosing and 553 on placebo [10]; 13 randomized trials [9]; 1,825 participants in 3 trials [11].
Sources: [10], [9], [11]
Caveats: The Cochrane review included 3 trials and pooled 2. For chronic low back pain it found only very low-quality evidence, from one trial that was later retracted [11].

CreatineStrong

Bone density

Not shown to change bone density in postmenopausal women or in older adults who exercise.

The numbers and sources

Effect: Bone mineral density did not change overall versus placebo in postmenopausal women [3]. In 20 trials (1,093 older adults, 69% women) of creatine with exercise training, total body bone mineral density did not change (MD 0.009 g/cm², p 0.557) [14].
Population: Postmenopausal women, 7 RCTs, 12 to 104 weeks [3]; older adults in exercise programs [14].
Sources: [3], [14]
Caveats: Each review had few bone trials. Neither found a bone effect.

GabapentinoidsStrong

Sciatica and low back pain

Not shown to reduce leg pain from sciatica or chronic low back pain against placebo, in one large trial and two meta-analyses.

The numbers and sources

Effect: In a trial of pregabalin for sciatica, leg pain at week 8 scored 3.7 on pregabalin and 3.1 on placebo (adjusted difference 0.5, 95% CI -0.2 to 1.2), and at week 52 it was 3.4 against 3.0. No secondary outcome differed, and there were 227 adverse events on pregabalin against 124 on placebo [8]. A meta-analysis found high-quality evidence of no effect of gabapentinoids on chronic low back pain in the short term (mean difference -0.0, 95% CI -0.8 to 0.7) or on lumbar radicular pain (mean difference -0.1, 95% CI -0.7 to 0.5) [9]. A second meta-analysis found gabapentin gave minimal improvement in chronic low back pain against placebo, and that pregabalin did worse than other analgesics [10].
Population: 209 patients with sciatica followed for a year [8]; 859 participants in 9 trials [9]; 8 trials, with 185 participants in the gabapentin comparison [10].
Sources: [8], [9], [10]
Caveats: Not an approved use. Evidence quality was very low in [10], and high for the main comparisons in [9].

Glucosamine and chondroitinStrong

Knee and hip osteoarthritis pain

Not shown to lower osteoarthritis pain by a clinically important amount: about 0.3 to 0.5 cm on a 10 cm scale, and independent trials found no difference.

The numbers and sources

Effect: Across 10 trials (3,803 patients), pain on a 10 cm scale differed from placebo by -0.4 cm for glucosamine (95% credible interval -0.7 to -0.1), -0.3 cm for chondroitin and -0.5 cm for the combination. None reached the prespecified minimal clinically important difference of -0.9 cm [1]. In 1,583 patients over 24 weeks, response rates did not differ significantly from placebo (60.1%): 3.9 points higher with glucosamine, 5.3 with chondroitin and 6.5 with both. Celecoxib was 10.0 points higher [2]. Individual patient data from 5 industry-independent trials (1,625 people) found glucosamine no different from placebo at 3 and 24 months, in every predefined subgroup [3].
Population: Adults with knee or hip osteoarthritis [1][2][3].
Sources: [1], [2], [3]
Caveats: Industry-independent trials showed smaller effects than commercially funded trials (P = 0.02 for interaction) [1]. In GAIT, an exploratory subgroup with moderate to severe pain responded more to the combination (79.2% versus 54.3%) [2].

Lymphatic drainage massageStrong

Swelling after knee replacement

Not shown to reduce leg swelling, pain or stiffness after knee replacement, in two meta-analyses of randomized trials.

Compared with no drainage [6]; no drainage or other physiotherapy [7]

The numbers and sources

Effect: Across 4 trials, knee flexion, pain and the circumference of the thigh, knee, calf and ankle did not differ with drainage [6]. Across 7 trials, knee flexion (SMD 0.03, 95% CI -0.22 to 0.28), knee extension, leg circumference (SMD -0.09, 95% CI -0.27 to 0.09) and pain (SMD -0.33, 95% CI -0.71 to 0.04) did not differ [7].
Population: 197 knee replacements, mean age 69.6, followed about 7 weeks [6]. 285 patients in 7 trials [7].
Sources: [6], [7]
Caveats: The two reviews share trials. Trials were small.

Vitamin DStrong

Fractures, falls and bone density

Not shown to lower fractures, hip fractures or falls in adults mostly not deficient or at high risk. Calcium with vitamin D had little to no effect.

The numbers and sources

Effect: Across 81 RCTs (53,537 participants), vitamin D had no effect on total fracture (RR 1.00, 95% CI 0.93 to 1.07), hip fracture (RR 1.11, 95% CI 0.97 to 1.26) or falls (RR 0.97, 95% CI 0.93 to 1.02). Results were similar for doses above 800 IU a day, and bone density differences ranged from -0.16% to 0.76% over 1 to 5 years [1]. A 2026 review of 69 trials (153,902 participants) found vitamin D alone did not change the risk of any fracture (RR 1.00, 95% CI 0.95 to 1.06, 36 trials, 92,045 participants, high certainty). Calcium with vitamin D gave RR 0.91 (95% CI 0.84 to 0.99, 15 trials), which the authors rated as little to no effect by absolute risk thresholds. Other fracture and fall outcomes showed little to no effect [2]. In VITAL, 2,000 IU a day for a median 5.3 years did not change total fractures (HR 0.98, 95% CI 0.89 to 1.08), nonvertebral fractures (HR 0.97) or hip fractures (HR 1.01, 95% CI 0.70 to 1.47) [3]. In the Finnish Vitamin D Trial, fractures were more frequent on 1,600 IU a day than placebo (HR 1.52, 95% CI 1.06 to 2.18) and did not differ on 3,200 IU a day (HR 1.07, 95% CI 0.72 to 1.58) over a mean 4.1 years [4]. Fall risk was 55% and fall-injury risk 11% in all three arms over 5 years [5].
Population: Adults in 81 trials [1]; adults not on osteoporosis drugs in 69 trials, 87% community dwelling and 73% not at high risk of fractures or falls [2]; 25,871 US men 50 and older and women 55 and older not selected for deficiency, low bone mass or osteoporosis [3]; 2,495 men 60 and older and women 65 and older in Finland, mean baseline 25-hydroxyvitamin D 75 nmol/L [4][5].
Sources: [1], [2], [3], [4], [5]
Caveats: Most participants were not selected for deficiency or high fracture risk, and evidence for high-risk people and residential care was limited [2]. The higher fracture count at 1,600 IU in the Finnish trial did not appear at 3,200 IU [4]. The bone density figures come from one review [1]. Trials that gave calcium with vitamin D were counted in [1] only where the groups differed by vitamin D alone.

AllopurinolModerate

Compared with febuxostat

Not shown to differ from febuxostat in gout flares or major cardiovascular events. One trial found fewer deaths on allopurinol.

The numbers and sources

Effect: Across 3 trials, gout attacks over 8 to 24 weeks occurred in 21% on allopurinol up to 300 mg and 23% on febuxostat 80 mg (RR 0.89, 95% CI 0.71 to 1.1) [1]. At fixed doses, 21% on allopurinol 300 mg reached the urate target against 53% and 62% on febuxostat 80 and 120 mg, with flares over weeks 9 to 52 in 64%, 64% and 70% [2]. With doses raised to target, flares in the last phase were 36.5% on allopurinol and 43.5% on febuxostat [4]. In people with gout and cardiovascular disease, the main cardiovascular outcome occurred in 10.4% on allopurinol and 10.8% on febuxostat (HR 1.03), and death from any cause was more frequent on febuxostat (HR 1.22, 95% CI 1.01 to 1.47), as was cardiovascular death (HR 1.34, 95% CI 1.03 to 1.73) [6].
Population: 1,136 participants in 3 trials [1]; 760 people with gout [2]; 940 people with gout [4]; 6,190 people with gout and cardiovascular disease followed a median of 32 months [6].
Sources: [1], [2], [4], [6]
Caveats: In the cardiovascular trial 56.6% stopped the study drug and 45.0% left follow-up [6].

BisphosphonatesModerateStudied in men

Fractures in men

Not shown to lower fractures in men over two years of use, because trials recorded too few fractures to tell. A Cochrane review called the effect very uncertain.

Compared with placebo [9]

The numbers and sources

Effect: Hip fractures occurred in 2 of 875 men on a bisphosphonate and 2 of 760 on placebo (RR 0.73, 95% CI 0.06 to 8.51). Symptomatic spine fractures were 5 of 1,021 against 7 of 855 (RR 0.49, 95% CI 0.14 to 1.74), and other fractures 25 of 1,130 against 19 of 913 (RR 0.78, 95% CI 0.42 to 1.45) [9].
Population: 2,548 men aged 50 or older in 7 placebo-controlled bisphosphonate trials, average age 52 to 73 [9].
Sources: [9]
Caveats: Very low certainty evidence, with few events and possible selective reporting [9]. The FDA approvals for men rest on bone density gains, not fracture reduction.

Calcium levelModerate

Albumin-corrected calcium against ionized calcium

Not shown to match ionized calcium better than plain total calcium: in two laboratory studies of 6,549 and 5,055 samples, correction formulas added no accuracy.

The numbers and sources

Compared with: Ionized (free) calcium measured in the same blood sample
Effect: In Norwegian laboratory data, unadjusted total calcium classified calcium status more accurately than albumin-adjusted calcium in people with reduced kidney function (Harrell's c 0.801 against 0.790 for locally derived formulas and 0.791 for the top-performing formula taken from the literature). With kidney function above that level the three did not differ [8]. In a second laboratory series, calcium adjusted by the Payne formula overestimated calcium status in up to 80% of cases with low albumin or a blood pH disorder. Three other formulas improved on the Payne formula but not on total calcium [9]. The corrected figure also depends on how albumin is measured: on 165 samples, mean corrected calcium was 2.34 mmol/L with one albumin method and 2.46 mmol/L with another [10].
Population: 6,549 patients older than 2 with ionized calcium, total calcium, albumin, creatinine and phosphate measured in one blood draw [8]. 5,055 samples with ionized calcium, total calcium, albumin and pH measured together [9]. 165 serum samples [10].
Sources: [8], [9], [10]
Caveats: Both large studies used samples on which ionized calcium had been ordered, so they describe hospital and clinic patients more than the general population. The authors of the first propose abandoning adjustment formulas, and the authors of the second conclude that total and albumin-adjusted calcium both misclassify calcium status [8][9].

CuppingModerate

Chronic pain, compared with sham cupping

Not shown to ease chronic pain more than sham cupping, in two meta-analyses and three later trials of back pain, knee osteoarthritis and fibromyalgia.

Compared with sham cupping [2][3][4][5][6]

The numbers and sources

Effect: Against sham cupping, pain did not differ significantly (SMD -0.27, 95% CI -0.58 to 0.05) and neither did disability (SMD -0.26) [2]. In 5 trials of dry against placebo cupping, the short-term pain difference was -9.9 points on a 100-point scale (95% CI -30.5 to 10.7), very uncertain at every time point [3]. In 90 people with chronic low back pain, pain differed by 0.0 to 0.6 points on a 0 to 10 scale over 8 weeks [4]. In 62 women with knee osteoarthritis, no outcome differed from sham at any time point [5]. In 141 people with fibromyalgia, pain at day 18 differed from sham by -3.0 points (95% CI -9.9 to 3.9) [6].
Population: 1,172 people with chronic pain in 18 trials [2]. 281 adults with musculoskeletal complaints in 5 trials [3]. 90 adults [4]. 62 women aged 50 to 75 [5]. 141 patients, 139 of them women [6].
Sources: [2], [3], [4], [5], [6]
Caveats: Few sham-controlled trials exist and most are small. The second meta-analysis rated certainty very low and four of its trials at high risk of bias [3].

Low-dose naltrexoneModerate

Chronic pain across conditions

Not shown to improve pain or quality of life across chronic pain conditions, with less pain than placebo only in fibromyalgia in one analysis.

The numbers and sources

Effect: In a meta-analysis of 7 RCTs, pain response did not differ from control overall (d -0.11, 95% CI -0.96 to 0.74). Against placebo in fibromyalgia, pain was lower (d -0.34, 95% CI -0.62 to -0.06). Against active drugs, there was no difference [6]. A systematic review of 7 RCTs (406 patients) at 2 to 4.5 mg a day found no improvement in pain or quality of life across chronic pain conditions [7]. In 23 people with osteoarthritis or inflammatory arthritis, 8 weeks of 4.5 mg did not change pain interference (-23 points against -22 on placebo, p 0.90), pain severity or fatigue [8].
Population: Adults with chronic pain conditions, including fibromyalgia [6][7], and 17 people with osteoarthritis and 6 with inflammatory arthritis [8].
Sources: [6], [7], [8]
Caveats: The arthritis trial was a pilot too small to rule out a modest effect [8].

Muscle relaxantsModerate

Added to an NSAID for acute low back pain

Not shown to improve function or pain at 1 week when added to naproxen or ibuprofen for acute low back pain in two emergency department trials.

The numbers and sources

Effect: In people sent home from an emergency department with naproxen, function at 1 week improved 10.1 points on a 24-point scale with cyclobenzaprine added and 9.8 points with placebo added (difference 0.3, 98.3% CI -2.6 to 3.2) [5]. In people sent home with ibuprofen, function improved 11.1 points with placebo added, 10.6 with baclofen, 10.1 with metaxalone and 11.2 with tizanidine. Moderate to severe pain at 1 week was reported by 30% on placebo, 33% on baclofen and 33% on tizanidine [6].
Population: 323 adults with acute low back pain of 2 weeks or less, 108 assigned cyclobenzaprine 5 mg [5]; 320 adults with acute low back pain of 2 weeks or less in 4 groups [6].
Sources: [5], [6]
Caveats: Both trials came from emergency departments in one city and followed people for 1 week as the main outcome.

21 more

Opioid pain relieversModerate

Acute low back, neck and limb pain

Not shown to relieve acute low back, neck or limb pain more than placebo or non-opioid pain relievers in three randomized trials.

The numbers and sources

Effect: In adults with acute low back or neck pain, pain at 6 weeks was 2.78 on a 10-point scale on oxycodone with naloxone and 2.25 on placebo (adjusted difference 0.53, 95% CI -0.00 to 1.07). Constipation was reported by 7.5% and 3.5% [3]. In emergency department patients with acute low back pain who were all given naproxen, adding oxycodone with acetaminophen did not improve function at 1 week more than adding placebo (difference 1.3 points on a 24-point scale, 98.3% CI -1.5 to 4.1) [4]. In emergency department patients with acute arm or leg pain, pain fell at 2 hours by 4.3 points of 10 on ibuprofen plus acetaminophen, 4.4 on oxycodone plus acetaminophen, 3.5 on hydrocodone plus acetaminophen and 3.9 on codeine plus acetaminophen, with no significant difference [5].
Population: 347 adults with 12 weeks or less of low back or neck pain, treated for up to 6 weeks [3]; 323 adults with acute low back pain, 108 of them assigned oxycodone with acetaminophen [4]; 416 adults aged 21 to 64 with moderate to severe acute extremity pain [5].
Sources: [3], [4], [5]
Caveats: The limb pain trial gave one dose and did not assess adverse events [5]. The back and neck pain trial used oxycodone combined with naloxone [3].

Oral corticosteroidsModerate

Gout flares, compared with NSAIDs

Not shown to differ from indomethacin or naproxen in pain relief for a gout flare in two randomized trials.

The numbers and sources

Effect: In emergency departments, pain fell by about 25 mm at rest and 45 mm with activity on a 100 mm scale over 14 days on both prednisolone and indomethacin. Minor adverse events in the emergency department were less frequent on prednisolone (6% against 19%) and equal over 14 days (37% each) [7]. In primary care, pain at 90 hours had fallen 44.7 mm on prednisolone and 46.0 mm on naproxen (difference 1.3 mm, 95% CI -9.8 to 7.1), with similar adverse effects [8].
Population: 416 adults with acute gout in 4 Hong Kong emergency departments, 376 completing [7]; 120 primary care patients with crystal-confirmed gout, given prednisolone 35 mg or naproxen 500 mg twice a day for 5 days [8].
Sources: [7], [8]
Caveats: Neither trial had a placebo group. In the larger trial gout was usually diagnosed on clinical criteria [7].

Osteopathy and osteopathic manipulative treatmentModerate

Cranial osteopathy

Not shown to rest on a reliable diagnosis or to have strong trial evidence. A review found examiners did not agree and rigorous efficacy trials were almost non-existent.

Compared with control conditions in randomized or crossover trials, which the review abstract does not describe [15]

The numbers and sources

Effect: Studies comparing two examiners, or one examiner on two occasions, found the diagnostic procedures of cranial osteopathy unreliable in many ways. Of 14 efficacy trials, 2 were at high risk of bias, 9 had major doubts about bias and 3 were at low risk, and those 3 did not consider other explanations such as a nonspecific effect of treatment [15].
Population: 9 reliability studies and 14 efficacy trials across many conditions [15].
Sources: [15]
Caveats: The review included only studies in English or French with open full text [15].

PRP injectionsModerate

Tennis elbow

Not shown to ease tennis elbow pain or function more than placebo injection at 3, 6 or 12 months, for PRP or whole-blood injections, in a Cochrane review.

Compared with placebo injection [6]

The numbers and sources

Effect: At 3 months, mean pain was 3.7 out of 10 with placebo and 0.16 points better with autologous blood or PRP (95% CI 0.60 better to 0.29 worse, 8 studies, 523 participants). Function was 1.86 points better on a 100-point scale (95% CI 4.9 better to 1.25 worse), and treatment success was 67% against 65% (RR 1.00, 95% CI 0.83 to 1.19). No clinically important benefit appeared at 6 or 12 months [6].
Population: 32 trials with 2,337 people with lateral elbow pain, 10 of them placebo-controlled [6].
Sources: [6]
Caveats: Evidence was moderate in certainty, downgraded for bias [6].

PRP injectionsModerate

Achilles tendinopathy

Not shown to improve chronic Achilles tendinopathy more than a sham or saline injection, in three placebo-controlled trials through one year.

Compared with sham injection, a dry needle under the skin not entering the tendon [7]; saline injection, both groups doing eccentric exercises [8][9]

The numbers and sources

Effect: At 6 months, VISA-A scores were 54.4 with PRP and 53.4 with sham (adjusted difference -2.7, 95% CI -8.8 to 3.3, clinically important difference 12) [7]. With eccentric exercise, VISA-A improved 21.7 points with PRP and 20.5 with saline at 24 weeks (difference -0.9, 95% CI -12.4 to 10.6) [8]. At one year the difference was 5.5 points (95% CI -4.9 to 15.8), and 59% in both groups were satisfied [9].
Population: 240 adults at 24 UK sites, mean age 52 [7]. 54 adults aged 18 to 70 in the Netherlands, followed for 24 weeks [8] and one year [9].
Sources: [7], [8], [9]
Caveats: The two smaller reports are the same trial at two time points [8][9].

SAM-eModerate

Osteoarthritis pain and function

Not shown to lower pain or improve function more than placebo in adults with knee or hip osteoarthritis.

The numbers and sources

Effect: In 4 trials (656 patients), pain was slightly lower than placebo, but not significantly (SMD -0.17, 95% CI -0.34 to 0.01), equal to 0.4 cm on a 10 cm scale. Function did not differ (SMD 0.02, 95% CI -0.68 to 0.71) [6].
Population: Adults with knee or hip osteoarthritis [6].
Sources: [6]
Caveats: The review rated trial quality and reporting as poor and called itself inconclusive [6].

YogaModerate

Chronic low back pain, against other exercise

Not shown to improve back function or pain more than other back exercise, stretching classes or physical therapy. It was noninferior to physical therapy in one trial.

Compared with back-focused exercise such as physical therapy [1]; conventional stretching classes [2]; 15 physical therapy visits [3]

The numbers and sources

Effect: Against back-focused exercise at 3 months, back function differed by 0.38 points (95% CI -1.33 to 0.62, 4 trials, moderate certainty) and pain by 2.68 points on a 0 to 100 scale (95% CI -2.01 to 7.36, very low certainty) [1]. Yoga was not superior to conventional stretching exercises at any time point [2]. Yoga met the noninferiority margins against physical therapy for function and pain at 12 weeks, and both were maintained at 1 year [3].
Population: 912 participants in 8 trials [1]. 183 adults in the yoga and stretching arms [2]. 320 adults [3].
Sources: [1], [2], [3]
Caveats: Participants were not blinded, and the physical therapy group had more loss to follow-up [3].

Alkaline phosphatase (ALP)LimitedStudied in children

Growth in children

Limited evidence: values in growing children run well above adult limits, and one series described temporary rises to 805 to 8,619 U/L lasting a mean 12 weeks.

The numbers and sources

Compared with: Adult reference limits
Effect: A German study built sex-specific percentile charts from birth to 18 years from 361,405 samples, and its authors write that interpretation in children is challenging because of extensive changes with growth and puberty [14]. The pediatric upper limits printed by one US survey manual, 321 to 382 U/L, are about 3 times its limit of 113 U/L for people older than 16 [4]. In 60 children diagnosed with transient hyperphosphatasemia, alkaline phosphatase ranged from 805 to 8,619 U/L (mean 2,311), the rise lasted a mean of 12 weeks, and the bone isoenzyme was elevated in most of the children in whom isoenzymes were measured [15].
Population: 124,440 patients from birth to 18 years at 6 German tertiary care centers and one laboratory service [14]. 60 children in Israel and 122 matched controls [15].
Sources: [14], [4], [15]
Caveats: The German charts were built from samples taken during clinical care, and the abstract prints no values [14]. The series of 60 children was retrospective and found no cause for the temporary rise [15].

AmitriptylineLimited

Chronic low back pain

One trial found low-dose amitriptyline did not reduce back pain more than an active placebo at 6 months, with less disability at 3 months only.

Compared with an active placebo, benztropine [11]

The numbers and sources

Effect: Pain did not differ at 6 months (adjusted difference -7.81, 95% CI -15.7 to 0.10) or 3 months. Disability was lower at 3 months (adjusted difference -1.62, 95% CI -2.88 to -0.36) but not at 6 months. Withdrawals for adverse events were 12% in each group [11].
Population: 146 adults with chronic nonspecific low back pain, 118 completing 6 months [11].
Sources: [11]
Caveats: One modest-sized trial at 25 mg a day [11].

BenzodiazepinesLimited

Acute low back pain

One trial found diazepam added to naproxen did not improve acute low back pain or function more than placebo at one week or three months.

The numbers and sources

Effect: Disability scores improved by 11 points in both the diazepam and placebo groups at one week. Moderate or severe pain at one week was reported by 32% on diazepam and 22% on placebo, and at three months by 12% and 9%. Adverse events were reported by 21% on diazepam and 15% on placebo [17].
Population: 114 emergency department patients with acute low back pain, all given naproxen [17].
Sources: [17], [8]
Caveats: One trial of 114 people. The diazepam label lists muscle spasm as an add-on use [8].

Calcium levelLimited

Surgery for mildly raised calcium from an overactive parathyroid

One trial found parathyroid surgery did not reduce deaths, fractures, heart events or kidney stones over 10 years in mild primary hyperparathyroidism.

The numbers and sources

Effect: In people with mild primary hyperparathyroidism and modest hypercalcemia, 15 had died 10 years after randomization, 8 assigned to surgery and 7 to observation. Over an extended period to 2018 there were 44 deaths, 24 and 20. Cardiovascular events, cerebrovascular events, cancer, peripheral fractures and kidney stones numbered 52 in the surgery group and 49 in the observation group. Vertebral fractures occurred in 7 patients in each group [7].
Population: 191 patients at 8 Scandinavian centers, 95 assigned to parathyroidectomy and 96 to observation [7].
Sources: [7]
Caveats: One trial of 191 people, and 50 withdrew during the study. It enrolled mild disease without known complications, and the abstract does not report calcium levels after surgery [7].

Deer antler velvetLimited

Arthritis

One trial found elk velvet antler did not differ from placebo over 6 months in 168 adults with rheumatoid arthritis.

The numbers and sources

Effect: In 168 people with rheumatoid arthritis on standard treatment, 6 months of elk velvet antler did not differ from placebo on joint pain and swelling, disease activity, function, quality of life or CRP [4]. A systematic review identified 2 rheumatoid arthritis trials and 1 osteoarthritis trial, and its authors wrote that the osteoarthritis findings may have some promise [5].
Population: Adults with stage 2 to 3 rheumatoid arthritis [4]; RCTs [5].
Sources: [4], [5]
Caveats: One osteoarthritis trial.

Dry needlingLimited

Knee osteoarthritis, compared with sham needling

Small trials found adding dry needling to exercise did not ease knee osteoarthritis pain or disability more than sham dry needling, through one year.

Compared with sham dry needling, both groups exercising [11][12]

The numbers and sources

Effect: At one year, pain differed from sham by 0.32 points (95% CI -1.12 to 1.18, P = 0.92) and WOMAC score by 0.29 points (95% CI -6.16 to 6.74). Both groups improved over time. In the dry needling group 90.3% reduced their medication, against 26.3% with sham [11]. In a 20-person pilot, pain and WOMAC scores improved in both groups with no group difference [12].
Population: 62 adults with knee osteoarthritis in a multicenter trial [11]. 20 adults aged 65 or older with thigh trigger points [12].
Sources: [11], [12]
Caveats: Two small trials from the same group, both adding needling to exercise.

MultivitaminLimited

Fractures

One trial found no drop in total, hip or nonvertebral fractures over 3.6 years in older US adults.

The numbers and sources

Effect: In COSMOS, Centrum Silver for a median 3.6 years did not lower total clinical fractures (aHR 1.09, 95% CI 1.00 to 1.19), hip fractures (aHR 1.06, 95% CI 0.80 to 1.42) or nonvertebral fractures (aHR 1.10, 95% CI 1.00 to 1.20) against placebo. 2,083 fractures occurred in all [10].
Population: 21,442 US women aged 65 and older and men aged 60 and older, not selected for osteoporosis [10].
Sources: [10]
Caveats: Fractures were self-reported. An ancillary analysis of a trial designed for cancer and cardiovascular outcomes.
Why limited: This rests on one trial, of 21,442 people. Strength counts independent studies that agree, so one trial is limited at any size.

Parathyroid hormone (PTH)Limited

Death and illness in a trial of surgery for mild primary hyperparathyroidism

One trial found parathyroid surgery for mild primary hyperparathyroidism did not change deaths, fractures or cardiovascular events over 10 years compared with observation.

The numbers and sources

Effect: After 10 years, 8 patients had died in the surgery group and 7 in the observation group. Over the extended follow-up to 2018 there were 24 and 20 deaths. Cardiovascular events, strokes, cancer, peripheral fractures and kidney stones together numbered 52 in the surgery group and 49 in the observation group, and 7 patients in each group had vertebral fractures [15].
Population: 191 patients with mild primary hyperparathyroidism at eight Scandinavian centers, 95 assigned to surgery and 96 to observation [15].
Sources: [15]
Caveats: One trial of 191 people, 50 of whom withdrew. It studied mild disease with modest high calcium and no known complications, so it does not describe more severe disease [15].

Physical therapy and exercise therapyLimited

Hip and knee osteoarthritis, compared with sham treatment

Two small trials found physical therapy programs did not ease hip or knee osteoarthritis pain more than sham ultrasound with inert gel.

Compared with sham ultrasound and inert gel [6][7]

The numbers and sources

Effect: In hip osteoarthritis, pain at week 13 fell from 58.8 to 40.1 mm with physical therapy and from 58.0 to 35.2 mm with sham, a difference of 6.9 mm favoring sham (95% CI -3.9 to 17.7). Function did not differ [6]. In knee osteoarthritis, pain fell by 2.2 cm with physiotherapy and 2.0 cm with placebo at 12 weeks, and global improvement was reported by 70% and 72% [7].
Population: 102 adults with painful hip osteoarthritis [6]. 140 adults with knee osteoarthritis, 119 completing [7].
Sources: [6], [7]
Caveats: The sham involved regular contact with a therapist, so both groups received attention. The authors of [7] wrote that the program did no more than regular contact with a therapist.

Tai chi and qigongLimited

Knee osteoarthritis, against physical therapy

One trial of 204 adults found tai chi and physical therapy eased knee osteoarthritis symptoms by similar amounts at 12 weeks, maintained to 52 weeks.

Compared with standard physical therapy twice a week for 6 weeks, then 6 weeks of monitored home exercise [10]

The numbers and sources

Effect: WOMAC scores fell 167 points with tai chi and 143 points with physical therapy at 12 weeks, a difference of 24 points (95% CI -10 to 58) that was not significant. Tai chi produced greater improvement in depression and in the physical component of quality of life [10].
Population: 204 adults with symptomatic knee osteoarthritis, mean age 60, 70% women [10].
Sources: [10]
Caveats: Patients knew their group. One urban academic hospital [10].

Tart cherryLimited

Knee osteoarthritis pain

One trial found knee pain scores did not differ significantly from placebo in adults with knee osteoarthritis, though CRP fell.

The numbers and sources

Effect: In 58 patients drinking two 8 oz bottles a day for 6 weeks, WOMAC scores fell during cherry juice but the difference from placebo was not significant (P = 0.16). hsCRP fell compared with placebo [14].
Population: Non-diabetic adults with Kellgren grade 2 to 3 knee osteoarthritis, crossover [14].
Sources: [14]
Caveats: One trial.

TurmericLimited

Hand osteoarthritis

One small pilot found hand pain fell from baseline over 3 months in adults over 50, but differences from placebo at 1 month were not significant.

The numbers and sources

Effect: In 27 adults over 50, 170 mg of curcumin a day for 3 months lowered VAS pain from baseline by 1.37 at rest and 1.74 during activity, and improved DASH scores by 11.84. Differences against placebo at 1 month were not significant [41].
Population: 27 adults with hand osteoarthritis, 13 on curcumin [41].
Sources: [41]
Caveats: A pilot trial. The month 3 changes are reported against baseline [41].

Uric acid (serum urate)Limited

During a gout flare

Pooled data from two trials found 14% of people in an acute gout flare had a serum urate of 6 mg/dL or below.

The numbers and sources

Compared with: Acute gout as diagnosed for entry to two treatment trials
Effect: Among 339 patients enrolled during an acute gout attack, 14% had a serum urate of 6 mg/dL or below at baseline and 32% had 8 mg/dL or below. Mean baseline serum urate was 7.1 mg/dL in people taking allopurinol and 8.5 mg/dL in those not taking it. The authors conclude that a normal level at presentation does not exclude an acute attack, and note that attacks occurred below 6.8 mg/dL, which they give as the saturation level for urate [22].
Population: 339 patients in 2 randomized trials of acute gout treatment, 94% men, mean age 50.5 [22].
Sources: [22]
Caveats: A secondary analysis of trials designed to compare two anti-inflammatory drugs. It did not follow the same people between flares [22].

Whole-body cryotherapyLimited

Rheumatoid arthritis

One study found people with rheumatoid arthritis improved about equally with whole-body cryotherapy and with traditional rehabilitation over 2 weeks.

The numbers and sources

Effect: Pain, disease activity, fatigue and walking measures improved similarly in both groups, as did interleukin-6 and TNF-alpha. The Health Assessment Questionnaire result was better in the traditional rehabilitation group (p < 0.05) [19].
Population: 25 patients given cryotherapy at -110°C and 19 given traditional rehabilitation [19].
Sources: [19]
Caveats: A comparison of two active programs, not described as randomized, with no untreated group.

Checked, no increase found1

H2 blockersModerate

Fractures

Not linked to a significant rise in hip fractures in a meta-analysis of observational studies, unlike proton pump inhibitors.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: H2 blocker use was not significantly associated with hip fracture (RR 1.12, 95% CI 0.97 to 1.30), while proton pump inhibitor use was (RR 1.30) [19].
Population: 11 observational studies, mostly of older adults [19].
Sources: [19]
Caveats: Observational. Fewer studies reported H2 blocker results than proton pump inhibitor results.

Not studied in people3

BPC-157Grey marketUntested

Tendon and ligament healing

Not studied in people. Animal studies found faster healing of cut tendons and ligaments in rats, from a handful of research groups.

No study in people. This rests on animal studies, lab studies.

The numbers and sources

Effect: A 2019 review found that studies of BPC-157 reported consistent healing effects on tendon, ligament and skeletal muscle, mostly in small rodent models, from a handful of research groups [15]. A 2025 systematic review included 35 preclinical studies and found that BPC-157 improved functional, structural and biomechanical outcomes in tendon, ligament, muscle and bone injuries in those models [16]. After the medial collateral ligament of the knee was cut in rats, BPC-157 given by injection, in drinking water or as a cream improved functional, biomechanical, macroscopic and microscopic healing over 90 days [17]. In the lab, it sped the outgrowth of fibroblasts from rat Achilles tendon tissue, raised their survival under oxidative stress and increased their migration, without directly changing how fast they multiplied [18]. In a 2026 study, rats given BPC-157 after Achilles tendon repair had a higher load to failure and lower tendon damage scores than controls at 4 weeks, but neither difference was statistically significant [19].
Population: Rats, and tendon cells from rats in the lab [15][16][17][18]. 32 rats in the 2026 study [19].
Sources: [15], [16], [17], [18], [19]
Caveats: The 2019 review notes that efficacy has not been confirmed in humans [15]. Surgically cut tendons and ligaments in rats do not show what happens in people. A 2026 review found no controlled human orthopaedic evidence for BPC-157 [10].

MOTS-cGrey marketUntested

Bone health

Not studied in people. In mice with their ovaries removed, MOTS-c reduced bone loss, and in the lab it slowed the formation of bone-resorbing cells.

No study in people. This rests on animal studies, lab studies.

The numbers and sources

Effect: In mice with ovariectomy-induced osteoporosis, MOTS-c treatment reduced bone loss measured by micro-CT. In the lab, it inhibited the formation of osteoclasts, the cells that break down bone, partly through AMPK [15]. A 2023 review reports that MOTS-c promotes the growth, maturation and mineralization of osteoblasts, the cells that build bone, and inhibits osteoclast production [16].
Population: Mice, and bone cells in the lab [15][16].
Sources: [15], [16]
Caveats: Bone loss after ovary removal in mice does not show what happens in people. FDA found no human exposure data for MOTS-c by any route [3].

TB-500Grey marketUntested

Tendon healing

Not studied in people. One rat study found stronger repaired Achilles tendons and better tissue scores after 4 weeks of TB-500.

No study in people. This rests on animal studies.

The numbers and sources

Effect: In rats whose Achilles tendon was cut and repaired, those given TB-500 for 4 weeks had a higher maximum load to failure than controls, a statistically significant difference, and lower total tendon damage scores on two microscopic scoring scales. Adding BPC-157 to TB-500 gave no further benefit over either alone [15]. A 2026 review found that direct musculoskeletal evidence for full-length thymosin beta-4 is limited to lab and rodent studies, and that TB-500 has no validated equivalence with it [9].
Population: 32 male rats, 8 in each of four groups [15].
Sources: [15], [9]
Caveats: One small study, which its authors describe as exploratory and preliminary [15]. Surgically cut tendons in rats do not show what happens in people. FDA found no clinical studies of TB-500 by any route [3].

Not yet researched3

CJC-1295 and ipamorelinGrey marketNo research yet