Osteoarthritis statistics — quick reference¶
Last checked: 2026-08-31
Figures below are not pooled across incompatible definitions. Each row preserves source, year, population and method. Confidence intervals are reproduced from the live-retrieved PubMed abstract; “UI” denotes GBD uncertainty interval.
Global burden¶
| Measure | Estimate | Population / year | Method | Source |
|---|---|---|---|---|
| People living with OA | 595 million (95% UI 535–656m) | Global, 2020 | GBD modeled symptomatic radiographic OA, 204 countries/territories | GBD 2021 OA Collaborators 2023, PMID 37675071 |
| Share of global population | 7.6% (95% UI 6.8–8.4) | Global, 2020 | Same | PMID 37675071 |
| Growth in total cases since 1990 | 132.2% (95% UI 130.3–134.1) | Global, 1990–2020 | Same | PMID 37675071 |
| Age-standardized YLD rate | 255.0 per 100,000 (95% UI 119.7–557.2) | Global, 2020 | GBD disability weights × severity prevalence | PMID 37675071 |
| YLD-rate increase | 9.5% (95% UI 8.6–10.1) | Global, 1990–2020 | Age-standardized | PMID 37675071 |
| Rank among causes of YLD, age ≥70 | 7th | Global, 2020 | GBD cause ranking | PMID 37675071 |
| High-BMI attributable fraction | 20.4% (95% UI -1.7 to 36.6) | Global, 2020 | Comparative risk assessment | PMID 37675071 |
Projected case growth, 2020–2050¶
| Site | Projected increase | Population | Method | Source |
|---|---|---|---|---|
| Knee | 74.9% (95% UI 59.4–89.9) | Global | Mixed-effects forecast | PMID 37675071 |
| Hand | 48.6% (35.9–67.1) | Global | Same | PMID 37675071 |
| Hip | 78.6% (57.7–105.3) | Global | Same | PMID 37675071 |
| Other OA | 95.1% (68.1–135.0) | Global | Same | PMID 37675071 |
Lifetime risk and national snapshots (added 2026-08-31)¶
| Measure | Estimate | Population / method | Source |
|---|---|---|---|
| Lifetime diagnosed symptomatic knee OA | 13.83%; 9.60% nonobese men to 23.87% obese women; median diagnosis age 55 | US NHIS 2007–08 + OAPol duration model | Losina 2013, PMID 23203864 |
| Incidence, ages 55–64 | 0.37%/year nonobese men to 1.02%/year obese women | Same | PMID 23203864 |
| Diagnosed by age 60 | 9.29% of US population | Same | PMID 23203864 |
| Radiographic knee OA, age ≥60 | 37.4%; symptomatic radiographic 12.1%; women 42.1% vs men 31.2% radiographic | NHANES III | Dillon 2006, PMID 17013996 |
| KL 3–4, age ≥60 | 12.9% women vs 6.5% men | NHANES III | PMID 17013996 |
| Knee replacement prevalence | 1.6% of US adults ≥60 | NHANES III | PMID 17013996 |
| Lifetime symptomatic hip OA by age 85 | 25.3% (95% CI 21.3–29.3) | Johnston County, n=3,068 | Murphy 2010, PMID 20713163 |
| Postindustrial vs early-industrial skeletal knee OA | 16% vs 6%; 2.1-fold (1.5–3.1) after age/BMI adjustment | US cadaver series | Wallace 2017, PMID 28808025 |
| GBD 2017 age-standardized prevalence | 3,754.2 per 100,000 (3,389.4–4,187.6) | 195 countries | Safiri 2020, PMID 32398285 |
| GBD 2017 incidence | 181.2 per 100,000 (162.6–202.4) | Same | PMID 32398285 |
| GBD 2010 knee / hip prevalence | 3.8% (3.6–4.1) / 0.85% (0.74–1.02) | GBD 2010 | Cross 2014, PMID 24553908 |
Hip-OA prevalence — estimates must retain definition¶
| Estimate | Population | Definition / method | Source year | Source |
|---|---|---|---|---|
| 8.55% (95% CI 4.85–13.18) | 326,463 participants, 31 studies worldwide | Pooled KL grade ≥2 | 2023 | Fan et al., PMID 36991481 |
| 1.20% (0.40–2.38) | African studies | Regional pooled KL-based estimate | 2023 | PMID 36991481 |
| 4.26% (0.02–14.93) | Asian studies | Same | 2023 | PMID 36991481 |
| 7.95% (1.98–17.36) | North American studies | Same | 2023 | PMID 36991481 |
| 12.59% (7.17–19.25) | European studies | Same | 2023 | PMID 36991481 |
| 9.42% (4.81–15.34) | Men | Pooled sex subgroup | 2023 | PMID 36991481 |
| 7.94% (3.57–13.81) | Women | Pooled sex subgroup; no significant sex difference | 2023 | PMID 36991481 |
Pain–radiograph discordance¶
| Conditional proportion | Range across studies | Population / method | Source |
|---|---|---|---|
| People with knee pain who had radiographic OA | 15–76% | Systematic search; heterogeneous definitions | Bedson et al. 2008, PMID 18764949 |
| People with radiographic knee OA who had pain | 15–81% | Same | PMID 18764949 |
These ranges are not prevalence estimates to average: they show the consequence of different samples, radiographic thresholds and pain definitions.
Sensitization and neuropathic-like features¶
| Measure | Estimate | Population / method | Source |
|---|---|---|---|
| Pain sensitization prevalence | 20% (95% CI 16–26); I²=89% | 53 studies, 7,117 knee-OA participants | Previtali et al. 2022, PMID 35356833 |
| Local pressure-pain threshold difference | SMD -1.00 (95% CI -1.67 to -0.32) | Knee OA vs healthy controls | PMID 35356833 |
| Distant pressure-pain threshold difference | SMD -0.54 (-0.76 to -0.31) | Same | PMID 35356833 |
| Pressure-pain threshold difference | SMD -0.85 (CI -1.1 to -0.6) | Earlier QST meta-analysis, knee OA vs controls | Fingleton et al. 2015, PMID 25749012 |
| Possible neuropathic-like pain, PainDETECT | 40% (95% CI 32–48) | Knee OA; questionnaire meta-analysis | Zolio et al. 2021, PMID 33971205 |
| Probable neuropathic-like pain, PainDETECT | 20% (15–24) | Knee OA | PMID 33971205 |
| S-LANSS-positive | 32% (26–38) | Knee OA | PMID 33971205 |
| DN4-positive | 41% (24–59) | Knee OA | PMID 33971205 |
| CSI sensitization prevalence | 36% (12–59) | Knee OA | PMID 33971205 |
| Possible neuropathic-like pain | 29% (22–37) | Hip OA, PainDETECT | PMID 33971205 |
| Probable neuropathic-like pain | 9% (6–13) | Hip OA, PainDETECT | PMID 33971205 |
Questionnaire-positive “neuropathic-like” pain does not establish a nerve lesion; estimates depend strongly on instrument.
Risk estimates¶
| Exposure | Estimate | Comparator / population | Method | Source |
|---|---|---|---|---|
| BMI 25 kg/m² | RR 1.59 (95% CI 1.34–1.81) | Reference BMI 22.5 | Dose-response meta-analysis, 12 studies | Zhou et al. 2014, PMID 24990315 |
| BMI 30 kg/m² | RR 3.55 (2.51–5.11) | Same | Same | PMID 24990315 |
| BMI 35 kg/m² | RR 7.45 (4.19–13.13) | Same | Same | PMID 24990315 |
| Occupational loading overall | OR 1.61 (1.45–1.78); I²=83.6% | Exposed vs sedentary work; 526,343 people | Meta-analysis | McWilliams et al. 2011, PMID 21382500 |
| Occupational loading, cohorts | OR 1.38 (1.10–1.74) | Cohort designs | Subgroup meta-analysis | PMID 21382500 |
| Occupational loading, cross-sectional | OR 1.57 (1.37–1.81) | Cross-sectional designs | Same | PMID 21382500 |
| Occupational loading, case-control | OR 1.80 (1.48–2.19) | Case-control designs | Same; publication bias detected | PMID 21382500 |
| OA after ACL injury | OR 6.81 (5.70–8.13) | ACL injury vs no injury | Umbrella review/meta-analysis | Webster et al. 2022, PMID 33852440 |
| OA after ACL reconstruction | OR 7.7 (6.05–9.79) | Reconstruction history vs no injury | Same | PMID 33852440 |
| OA prevalence ≈10 years after ACL reconstruction | 36% (19.70–53.01) | Reconstructed cohorts | Same; high heterogeneity | PMID 33852440 |
Exercise effects — knee OA¶
| Outcome / time | Mean effect on 0–100 scale | Evidence base | Source |
|---|---|---|---|
| Pain, immediately after treatment | 12-point improvement (95% CI 10–15) | 44 land-exercise trials | Fransen et al. 2015, PMID 26405113 |
| Physical function, immediately | 10 points (8–13) | Same review | PMID 26405113 |
| Quality of life, immediately | 4 points (2–5) | 13 studies | PMID 26405113 |
| Pain, 2–6 months after formal treatment | 6 points (3–9) | 12 studies | PMID 26405113 |
| Function, 2–6 months after formal treatment | 3 points (1–5) | 10 studies | PMID 26405113 |
| Trials with adequate sequence generation, concealment and incomplete-data handling | 19/54 (35%) | Risk-of-bias assessment | PMID 26405113 |
Intensive diet and exercise (IDEA)¶
| Outcome at 18 months | Diet + exercise | Diet | Exercise | Between-group statistic | Source |
|---|---|---|---|---|---|
| Weight loss | 10.6 kg (11.4%) | 8.9 kg (9.5%) | 1.8 kg (2.0%) | Descriptive | Messier et al. 2013, PMID 24065013 |
| Knee compressive force | — | 2,487 N | 2,687 N | Exercise–diet 200 N (95% CI 55–345) | PMID 24065013 |
| WOMAC pain (0–20) | 3.6 | 4.8 | 4.7 | Exercise–combined 1.02 (0.33–1.71) | PMID 24065013 |
| WOMAC function (0–68) | 14.1 | 18.4 | 18.4 | Exercise–combined 4.29 (2.07–6.50) | PMID 24065013 |
| Physical HRQoL | 44.7 | — | 41.9 | Combined–exercise 2.81 (0.86–4.76) | PMID 24065013 |
| Retention | 399/454 (88%) | All groups | 18-month RCT | — | PMID 24065013 |
Semaglutide in obesity-associated knee OA (STEP 9)¶
| Outcome at week 68 | Semaglutide | Placebo | Contrast | Source |
|---|---|---|---|---|
| Body weight change | -13.7% | -3.2% | P<.001 | Bliddal et al. 2024, PMID 39476339 |
| WOMAC pain change | -41.7 | -27.5 | P<.001 | PMID 39476339 |
| SF-36 physical-function change | +12.0 | +6.5 | P<.001 | PMID 39476339 |
| Permanent discontinuation due to adverse events | 6.7% | 3.0% | GI disorders most common | PMID 39476339 |
| Baseline profile | mean age 56; BMI 40.3; WOMAC pain 70.9; 81.6% women | Trial population | Limits transfer beyond obesity-associated knee OA | PMID 39476339 |
Medicines¶
| Comparison / outcome | Estimate | Evidence base | Source |
|---|---|---|---|
| Topical NSAID vs acetaminophen, function | SMD -0.29 (95% CrI -0.52 to -0.06) | Network of 122 RCTs, 47,113 participants | Zeng et al. 2021, PMID 34174454 |
| Topical vs oral NSAID, function | SMD 0.03 (-0.16 to 0.22) | Same | PMID 34174454 |
| Topical NSAID vs oral NSAID, GI adverse effects | RR 0.46 (0.34–0.61) | RCT network | PMID 34174454 |
| Oral NSAID regimens with increased AE-withdrawal risk | 18.5% of evaluated regimens | Network meta-analysis | da Costa et al. 2021, PMID 34642179 |
| Opioid regimens with increased AE-withdrawal risk | 83.3% | Same | PMID 34642179 |
| Opioid regimens with increased any-AE risk | 89.5% | Same | PMID 34642179 |
| Tramadol pain absolute improvement | 4% (95% CI 3–5) | 8 trials, 3,972 participants | Toupin April et al. 2019, PMID 31132298 |
| Tramadol adverse events | RR 1.34 (1.24–1.46) | 4 studies, 2,039 participants | PMID 31132298 |
| Tramadol AE withdrawal | RR 2.64 (2.17–3.20) | 9 studies, 4,533 participants | PMID 31132298 |
| Tramadol serious adverse events | RR 1.78 (1.11–2.84) | 7 studies, 3,612 participants | PMID 31132298 |
| Duloxetine pain mean difference | -0.88 (95% CI -1.11 to -0.65) | 3 RCTs; 992 participants for endpoint | Wang et al. 2015, PMID 26176791 |
| Duloxetine ≥30% pain response | RR 1.49 (1.31–1.70) | 989 participants | PMID 26176791 |
| Duloxetine ≥50% pain response | RR 1.69 (1.27–2.25) | 989 participants | PMID 26176791 |
| Duloxetine adverse events | RR 2.15 (1.48–3.11) | 1,011 participants | PMID 26176791 |
Injections¶
| Intervention / outcome | Estimate | Population / method | Source |
|---|---|---|---|
| Viscosupplementation pain vs placebo | SMD -0.08 (95% CI -0.15 to -0.02) | 24 large placebo-controlled trials, 8,997 participants | Pereira et al. 2022, PMID 36333100 |
| Same translated to 100-mm VAS | -2.0 mm (-3.8 to -0.5) | Same | PMID 36333100 |
| Viscosupplementation serious adverse events | RR 1.49 (1.12–1.98) | 15 large trials, 6,462 participants | PMID 36333100 |
| RESTORE pain change | -2.1 PRP vs -1.8 saline | 288 participants, 12 months | Bennell et al. 2021, PMID 34812863 |
| RESTORE pain contrast | -0.4/10 (95% CI -0.9 to 0.2) | MCID 1.8 | PMID 34812863 |
| RESTORE cartilage-volume contrast | -0.2% (-1.9 to 1.5) | Medial tibial MRI volume | PMID 34812863 |
| RESTORE null secondary outcomes | 29/31 | Prespecified secondary outcomes | PMID 34812863 |
| Repeated triamcinolone cartilage change | -0.21 vs -0.10 mm | 140 participants; every 12 weeks, 2 years | McAlindon et al. 2017, PMID 28510679 |
| Triamcinolone cartilage contrast | -0.11 mm (-0.20 to -0.03) | MRI index compartment | PMID 28510679 |
| Triamcinolone pain contrast | -0.6/20 (-1.6 to 0.3) | WOMAC pain | PMID 28510679 |
| Hip injection network: active treatment vs saline | No intervention significantly superior at 2–4 or 6 months | 11 RCTs, 1,353 participants | Gazendam et al. 2021, PMID 32829298 |
Procedures and surgery¶
| Outcome | Estimate | Population / method | Source |
|---|---|---|---|
| Sham meniscectomy trial: Lysholm contrast | -1.6 (95% CI -7.2 to 4.0) | 146 people with degenerative tear, no OA; 12 months | Sihvonen et al. 2013, PMID 24369076 |
| WOMET contrast | -2.5 (-9.2 to 4.1) | Same | PMID 24369076 |
| Exercise-pain contrast | -0.1/10 (-0.9 to 0.7) | Same | PMID 24369076 |
| TKR strategy KOOS4 improvement | 32.5 vs 16.0 | 100 surgery-eligible knee-OA patients; 12 months | Skou et al. 2015, PMID 26488691 |
| TKR adjusted mean difference | 15.8 (95% CI 10.0–21.5) | TKR+nonsurgical vs nonsurgical | PMID 26488691 |
| Serious adverse events | 24 vs 6 (P=.005) | Same | PMID 26488691 |
| Nonsurgical-to-TKR crossover | 13/50 (26%) | By 12 months | PMID 26488691 |
| UKA vs TKA hospital stay, RCT group | -1.20 days (-1.67 to -0.73) | Systematic review/meta-analysis | Wilson et al. 2019, PMID 30792179 |
| UKA vs TKA 5-year revision RR, trials | 5.95 (1.29–27.59) | Same | PMID 30792179 |
| UKA vs TKA 5-year revision RR, registries | 2.50 (1.77–3.54) | Same | PMID 30792179 |
| UKA vs TKA 5-year revision RR, cohorts | 3.13 (1.89–5.17) | Same | PMID 30792179 |
Genetics¶
| Finding | Estimate | Population / method | Source |
|---|---|---|---|
| GWAS meta-analysis sample | 826,690 people; 177,517 OA | 9 populations, 11 phenotypes | Boer et al. 2021, PMID 34450027 |
| Independent risk variants | 100; 52 previously unreported | Same | PMID 34450027 |
| SMO rare missense variant and hip OA | OR 2.8; frequency 0.11%; P=7.9×10^-12 | Iceland + UK Biobank meta-analysis | Styrkarsdottir et al. 2018, PMID 30374069 |
| IL11 missense variant and hip OA | OR 1.30; frequency 2.08%; P=2.1×10^-11 | Same | PMID 30374069 |
| COL11A1 common missense variant | OR 1.08; frequency 61%; P=5.2×10^-10 | Same | PMID 30374069 |
| CHADL1 recessive association | OR 5.9; P=1.8×10^-25 | Same | PMID 30374069 |
Paracetamol, SPACE, glucosamine¶
| Comparison | Estimate | Evidence base | Source |
|---|---|---|---|
| Paracetamol vs placebo, hip/knee pain | WMD −3.7/100 (−5.5 to −1.9); not clinically important | High-quality GRADE | Machado 2015, PMID 25828856 |
| Cochrane paracetamol, pain | 3.23/100 better than placebo (5.43 to 1.02 better) | 10 RCTs | Leopoldino 2019, PMID 30801133 |
| Cochrane paracetamol, function | 2.9/100 better (0.95 to 4.89) | Same | PMID 30801133 |
| SPACE 12-month BPI interference | 3.4 opioid vs 3.3 nonopioid (diff 0.1, −0.5 to 0.7) | n=240 VA, 97.5% complete | Krebs 2018, PMID 29509867 |
| SPACE 12-month BPI severity | 4.0 vs 3.5 (diff 0.5, 0.0 to 1.0) favoring nonopioid | Same | PMID 29509867 |
| GAIT 20% pain-response vs placebo 60.1% | Glucosamine +3.9 pp (p=0.30); CS +5.3 (p=0.17); combo +6.5 (p=0.09); celecoxib +10.0 (p=0.008) | n=1,583 | Clegg 2006, PMID 16495392 |
| Wandel glucosamine/CS/combo vs placebo, 10 cm VAS | −0.4 (−0.7 to −0.1) / −0.3 (−0.7 to 0.0) / −0.5 (−0.9 to 0.0) cm; MCID −0.9 cm | 10 trials, 3,803 people | Wandel 2010, PMID 20847017 |
NSAID vascular harm¶
| Outcome | Estimate | Evidence base | Source |
|---|---|---|---|
| Coxib major vascular events | RR 1.37 (1.14–1.66) | 280 placebo trials, 124,513 people | CNT 2013, PMID 23726390 |
| Diclofenac major vascular events | RR 1.41 (1.12–1.78) | Same | PMID 23726390 |
| Naproxen major vascular events | RR 0.93 (0.69–1.27) | Same | PMID 23726390 |
| Extra major vascular events | ~3 per 1,000 patient-years on coxib or diclofenac, one fatal | Same | PMID 23726390 |
| Rofecoxib MI vs placebo | Rate ratio 2.12 (1.26–3.56) | 31 trials, 116,429 people | Trelle 2011, PMID 21224324 |
| Real-world MI, 1–7 days naproxen | OR 1.53 (1.07–2.33) | 446,763 people, 61,460 MIs | Bally 2017, PMID 28487435 |
| Real-world MI, 1–7 days diclofenac | OR 1.50 (1.06–2.04) | Same | PMID 28487435 |
Landmark procedure and DMOAD numbers (added 2026-08-31)¶
| Outcome | Estimate | Source |
|---|---|---|
| Moseley 1-year Knee-Specific Pain Scale | Sham 48.9, lavage 54.8, debridement 51.7 | Moseley 2002, PMID 12110735 |
| Kirkley 2-year WOMAC | Surgery 874 vs control 897 (diff −23, −208 to 161) | Kirkley 2008, PMID 18784099 |
| MeTeOR 6-month WOMAC function improvement | 20.9 vs 18.5 (diff 2.4, −1.8 to 6.5); 30% PT crossover | Katz 2013, PMID 23506518 |
| Kise 2-year KOOS4 | Diff 0.9 (−4.3 to 6.1); 19% exercise-to-surgery crossover | Kise 2016, PMID 27440192 |
| TOPKAT 5-year OKS | Diff 1.04 (−0.42 to 2.50); n=528 | Beard 2020, PMID 32369436 |
| Frydendal 6-month Oxford Hip Score | THR +15.9 vs training +4.5 (diff 11.4, 8.9–14.0); 21% crossover | Frydendal 2024, PMID 39476341 |
| FORWARD 2-year cartilage thickness, 100 µg q6mo vs placebo | +0.05 mm (0.03–0.07); no significant WOMAC difference | Hochberg 2019, PMID 31593273 |
| Lorecivivint OA-11 Pain NRS week 12 | −2.24 vs −2.49 (p=0.185); n=513 | Yazici 2025, PMID 39808286 |
| Tanezumab 2.5 / 5 mg composite joint-safety vs NSAID | 38.3 (28.0–52.5) / 71.5 (56.7–90.2) vs 14.8 (8.9–24.6) per 1,000 PY | Hochberg 2021, PMID 33538113 |
| Tanezumab pooled SC composite joint-safety | 145/4,541 (3.2%); 0% placebo, 3.2% 2.5 mg, 6.2% 5 mg, 1.5% NSAID | Carrino 2023, PMID 37652258 |
| Methotrexate vs placebo, 52-week VAS pain | Diff 0.3 mm (−6.7 to 7.3); n=215 | Zhu 2025, PMID 40455462 |
| Septic arthritis after IA GC | 0.08% (0.03–0.12) of 14,118 injections | Petersen 2019, PMID 31146626 |
| RPOA after hip CSI | 6% (3–9%) in studies able to detect it | Sabatini 2023, PMID 37941925 |
| GLA:D pain change | −12.4 at 3 months, −13.7 at 12 months (0–100); n=9,825 | Skou 2017, PMID 28173795 |
| Christensen weight-loss pain ES | 0.20 (0 to 0.39) at 6.1 kg (4.7–7.6 kg) | Christensen 2007, PMID 17204567 |
| Cam deformity → incident hip OA | OR 2.11 (1.55–2.87) | Saberi Hosnijeh 2017, PMID 27696746 |
| Acetabular dysplasia → incident hip OA | OR 2.19 (1.50–3.21) | PMID 27696746 |
| Framingham KL 0 MRI “any abnormality” | 89% (631/710); osteophytes 74%; cartilage 69%; BML 52% | Guermazi 2012, PMID 22932918 |
| Felson BML in painful vs painless radiographic OA | 77.5% vs 30%; large lesions 35.9% vs 2% | Felson 2001, PMID 11281736 |
| da Costa 2017 diclofenac 150 mg pain ES | −0.57 (−0.69 to −0.45); etoricoxib 60 mg −0.58 (−0.74 to −0.43) | da Costa 2017, PMID 28699595 |
| Rutjes large blinded HA trials | Pain ES −0.11 (−0.18 to −0.04); unpublished −0.03 (−0.14 to 0.09) | Rutjes 2012, PMID 22868835 |
| Hip precaution early dislocation | RCT RR 1.8 (0.6–5.2); NRS RR 0.9 (0.3–2.5); n=8,835 | Korfitsen 2023, PMID 37039064 |
| Doxycycline 30-month medial JSN | 0.30±0.60 vs 0.45±0.70 mm; no pain reduction | Brandt 2005, PMID 15986343 |
| ADAMTS5 knockout | First single-gene deletion to abrogate murine OA cartilage loss | Glasson 2005, PMID 15800624 |
Known conflicts and caveats¶
- GBD estimates are model outputs with uncertainty intervals; they are not a direct census.
- Radiographic, symptomatic and coded OA prevalence estimates answer different questions.
- Sensitization and neuropathic-like prevalence varies by questionnaire or QST method; instruments are not interchangeable diagnoses.
- Network meta-analysis rankings can exaggerate small differences and inherit heterogeneity across dose, comparator and risk of bias.
- Injection trials have large contextual effects; saline is not equivalent to no treatment.
- Surgery comparisons combine RCTs, registries and cohorts with different confounding structures; they are displayed separately rather than averaged.
- Relative risks can appear large when absolute event rates are low; where abstracts did not provide absolute rates, none were invented.
- GBD 2010, 2017 and 2021 prevalence estimates use expanding case definitions and must not be averaged.
- Wallace's 2.1-fold modern increase after age/BMI adjustment conflicts with the heuristic that OA growth is “just demography and obesity.”
- Murphy's null BMI–lifetime-hip-OA association conflicts with incidence studies linking obesity to hip OA.
- Naproxen is approximately null for major vascular events in CNT RCT IPD (RR 0.93) but not in Bally's real-world MI IPD (OR 1.53 at 1–7 days) — trial vs observational discrepancy, not a number to average.
- GAIT's overall glucosamine/CS null and moderate–severe subgroup positive result is the opposing pole to Wandel's industry-independent smaller effects.