Supplements
curcumin decreases osteoarthritis pain
In plain terms: Does turmeric help arthritis and joint pain?
Strong support Supplements 💰 Industry COI noted🔬 Includes disconfirming
Part of: 🧪 turmeric & curcumin
RefutedContestedStrong support
consensus score 0.91
Yes — this is curcumin's strongest claim. Multiple meta-analyses find standardized curcumin extract eases knee-osteoarthritis pain about as well as anti-inflammatory drugs over the short term. Note that's supplement doses, not turmeric sprinkled on food.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: All trials, pooled (Meta-analysis)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
9 support 0 contradict 3 tested null 0 mixed · 12 sources, 9 independent groups
What the evidence shows
Curcumin's best-supported human claim. Multiple independent meta-analyses of RCTs find standardized curcumin/turmeric extract (~1000 mg/day) reduces knee-osteoarthritis pain and improves function, roughly comparable to NSAIDs over short-term use — though trials are mostly short and use bioavailability-enhanced formulations, not dietary turmeric.
The evidence (23)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Chen J et al 2025 · Front Pharmacol | meta-analysis | supports | low | Full-text re-read 2026-08-25 (rubric v1.1.3 axis 2): UMBRELLA REVIEW, ROLE = review, ZERO WEIGHT. The units are SYNTHESES, not primary trials - the search set out 'identifying all systematic reviews (SRs) and meta-analyses (MAs) on curcumin treatment for KOA' and '7 SRs and MAs that concentrated on the therapeutic application of curcumin in the context of KOA were identified for inclusion in the analysis' (Onakpoya 2017, Dai 2021, Hsiao 2021, Feng 2022, Ji 2022, Yu 2025, Hidayat 2025). It also pools nothing of its own: 'we reframed our work as a critical overview rather than conducting a new quantitative umbrella meta-analysis... we did not provide updated pooled effect estimates beyond those already published in previous umbrella meta-analyses'. Both the old and the new wording of the rule therefore land in the same place. The authors give the double-count reason themselves - the reframing was 'to avoid duplicate counting of primary trials, which we identified as substantial (corrected covered area = 48.25%) using the GROOVE tool' - and that same duplication is what disqualifies it from counting beside this claim's own synthesis rows, which pool the same small curcumin-KOA RCT literature. DIRECTION IS UNCHANGED AND REAL on the claim's object (curcumin -> knee-OA pain, VAS and WOMAC-pain): 'Seven SRs met inclusion criteria. Curcumins demonstrated potential efficacy and safety advantages over control treatments in KOA management.' It is the weight, not the sign, that this re-read removes. DENOMINATORS: 7 SRs; per the GROOVE figure 19 index publications and 74 primary-study occurrences; 48 outcome estimates re-GRADEd, of which 'a total of 37 outcome measures had critically low quality evidence, 6 had low quality evidence, and only 5 had moderate quality evidence' and none high; AMSTAR 2 put every review at the bottom - 'All SRs were rated as having critically poor quality' - and ROBIS put 4 of 7 at overall high risk of bias. No pooled effect estimate exists in this paper, and none should be read off its Table 2: the sign conventions differ across the seven reviews (Yu 2025 and Hidayat 2025 report positive VAS mean differences for benefit, the other five negative), and two VAS pools cross the null ('VAS score 7 (757) MD: -1.66 (-3.37, 0.06)'; Hidayat 'VAS score 6 (448) MD: 18.25 (-7.79, 28.72)'). QUALITY STAYS low, but the previous rationale was wrong: 'critically poor quality' is this paper's verdict on the reviews it appraised, not on itself, and on its own design it is above average - PROSPERO CRD42025641801, two independent reviewers, eight databases including four Chinese-language ones, explicit overlap quantification. It earns low on its own reporting instead, for three internal contradictions in the text: the GRADE downgrade percentages are transposed between Results ('inconsistency in results was observed in 77.1% of the cases, publication bias was a factor in 45.8% of the cases, imprecision in the data was present in 29.2% of the cases') and Discussion ('Publication bias was prevalent, noted in 77.1% of cases, followed by issues of imprecision (45.8%) and inconsistency (29.2%).'); the corrected covered area is 48.25% in the text but 55.64% in Figure 5, with the CCA formula printed differently in Methods and Results; and a sentence about an entirely different intervention pair survives from another manuscript - 'the evidence quality for comparisons like PRP versus HA in treating KOA' - in a paper that studied neither. Grade left as filed at meta-analysis and FLAGGED: the paper ran no meta-analysis, but the vocabulary has no umbrella value and the zero weight is carried by evidence_role. Independence flags, not acted on: the source note stores 'Chen J et al' while the paper has two Chen J authors (Chen Jixin first, Chen Jianliang sixth), so that string cannot be matched safely; and Bideshki, first author of the sibling zero-weighted umbrella s38576215 on this same claim, is a named peer reviewer of this paper. |
| Toumi H et al 2026 · study_type: observational | observational | supports | moderate | Narrative review of 165 studies on curcumin in arthritis; states clinical evidence in OA indicates improvements in pain/function, efficacy sometimes comparable to NSAIDs, but calls evidence 'limited' and cautions on bioavailability/formulat |
| Raja K et al 2026 · Int J Mol Sci | in-vitro | supports | low | Full-text re-read 2026-08-25 (JATS full text, PMC12841081; Int J Mol Sci 2026;27(2):609). ROLE = off-scope on the OBJECT, ZERO WEIGHT, matching the sibling rulings already applied to s41303695 and s41590033 on this claim. IN VITRO ONLY: human BMSC-derived chondrocytes differentiated 21 d in MesenCult-ACF, then 'chondrocytes were treated with 10 µg/mL LPS for 24 h', then curcumin at 25 or 100 µM for 24 h; 'Data were presented as mean ± SD from n = 3 independent biological replicates'. The measured endpoints are Rhodamine-123 mitochondrial membrane potential, Acridine-Orange acidic vesicular organelles, propidium-iodide membrane integrity, and Western blots of pro-caspase-3, Bcl-2, Beclin-1 and LC3-II normalised to GAPDH. No pain, function or behavioural outcome is measured, and there is no organism to measure one in - 'The experimental procedures in this research were carried out in vitro and did not entail the utilization of animal models, patient samples, nor human subjects.' DIRECTION AND DENOMINATORS, corrected from the previous extract: the effects are favourable but NON-MONOTONIC and figure-only, with no numeric effect estimate reported anywhere in the text for any endpoint. 'Curcumin treatment upregulated Beclin-1 and LC3-II levels at 25 µM but showed a moderate decrease at 100 µM'. Apoptosis inhibition is inferred from preservation of the zymogen band, not from loss of an active fragment - 'treatment with curcumin at 25 µM and 100 µM restored the pro-caspase-3 band intensity, suggesting that curcumin inhibited caspase-3 cleavage and thereby attenuated apoptotic activation'. The mitochondrial result is displayed for the 100 µM arm alone (Figure 2C: '( C ). Chondrocytes treated with 100 µM Curcumin in the presence of LPS.'), so the abstract's 'Curcumin (25 µM and 100 µM) restored ΔΨm, increased Beclin-1 and LC3-II, and maintained pro-caspase-3 expression, with Bcl-2 showing a non-monotonic response (higher at 25 µM than at 100 µM).' overstates that figure's own contents; the previous extract's '25-100uM' range and flat 'reduced apoptotic signaling' inherited that overstatement. The Results also record the LPS insult itself as NULL on autophagy - 'LPS exposure did not significantly alter the Beclin-1 or LC3-II expression, indicating that basal autophagic activity was largely maintained under inflammatory conditions.' - contradicting this paper's own abstract and discussion. The authors state the translational ceiling themselves: 'However, the study is limited by its in vitro design, which may not fully replicate the complex in vivo joint environment. Additionally, curcumin’s poor bioavailability in physiological conditions was not addressed, which could affect translational relevance.' |
| Lopresti AL et al 2026 · study_type: RCT | RCT | supports | moderate | RCT, n=60 adults with knee OA, topical curcumin gel (VAS-101) vs placebo gel over 28 days; KOOS pain score improved (beta 5.12, 95%CI 0.47-9.77, d=0.62, p=0.041) and daily pain ratings improved (p=0.005) vs placebo. |
| Bideshki 2024 · Phytother Res | meta-analysis | supports | moderate | Umbrella meta-analysis: the pooled UNITS are published syntheses, not trials — 'PubMed, Scopus, Embase, and Web of Science databases were employed to find English-language meta-analyses of randomized controlled trials investigating the effect of curcumin supplementation on OA outcomes up to September 2023', and 'The pooled effect of the 11 included meta-analyses' is the whole result. ROLE = review, ZERO WEIGHT (rubric v1.1.3 axis 2, from reread #690): a random-effects pool OF meta-analyses carries no independent observation and cannot count beside the syntheses it re-passes; its search window to September 2023 spans this claim's own counting synthesis rows (s27533649 2016, s29018060 2017, s34017975 2021, plus the already-duplicate-population s35935936), so the same primary trials would vote a third time. Which 11 were included cannot be verified from the abstract. DIRECTION IS UNCHANGED AND REAL on the claim's object: 'The pooled effect of the 11 included meta-analyses showed that curcumin could significantly decrease the VAS score (weighted mean difference [WMD] and standardized mean difference [SMD]), WOMAC-total (SMD and WMD), WOMAC-Function (SMD and WMD), WOMAC-Pain (SMD), and WOMAC-Stiffness scores (SMD) (p ≤ 0.001, ≤0.001, ≤0.001, 0.007, ≤0.001, 0.002, ≤0.001, ≤0.001, respectively)'. DENOMINATOR/ENDPOINT: n = 11 meta-analyses; no trial count, no participant count, and — for all eight pooled estimates — no effect size and no confidence interval are reported, only p-values. WOMAC-Pain and WOMAC-Stiffness reach significance on SMD only, with no WMD companion, i.e. no magnitude on the natural pain scale. QUALITY moderate, not high: the authors' conclusion outruns their own stated analysis set — 'The results strongly support curcuminoid supplementation in relieving pain, improving joint mobility and stiffness, and shortening medication usage of OA patients', while the endpoints named are only 'The visual analog scale (VAS), Western Ontario and McMaster Universities Arthritis Index (WOMAC) pain, function, and stiffness scales'; no medication-use analysis is described anywhere in the available text. Publication-bias testing, AMSTAR/GRADE appraisal of the 11 reviews, and any correction for primary-trial overlap between them are cannot-tell from the abstract — not established either way. CORRECTION to the previous extract: 'despite between-review heterogeneity' is unsupported; the abstract says only that 'Heterogeneity was determined by I2 statistics and the Cochrane Q-test', reporting no I2 value. Abstract-grade and permanently so — Europe PMC reports isOpenAccess=N, inEPMC=N, no PMCID. |
| Lan CW et al 2026 · study_type: animal | animal | tested-null | low | Rat MIA-induced knee OA model; turmeric fermentation liquid reduced inflammatory cytokines (TNF-a, IL-1b), increased IL-10, and reduced pain behavior and cartilage degradation vs MIA control. |
| Liu X, Machado GC, Eyles JP, Ravi V, Hunter DJ 2018 · Br J Sports Med | meta-analysis | supports | moderate | Systematic review with random-effects meta-analysis and meta-regression (NOT a network meta-analysis, as previously recorded here): 'An intervention systematic review with random effects meta-analysis and meta-regression', 69 eligible RCTs of 20 oral supplements vs placebo in hand, hip or knee OA. Both 'curcumin' and 'Curcuma longa extract' are named among the seven supplements that 'demonstrated large (effect size >0.80) and clinically important effects for pain reduction at short term'; 'Similar results were found for physical function.' DENOMINATOR: the 69 studies span all 20 supplements - the abstract reports no per-supplement trial count, participant count or effect estimate, so the curcumin subset's own denominator is cannot-tell from the available text. DURATION: short term only - 'No supplements were identified with clinically important effects on pain reduction at long term.' Quality moderate, not high: the authors' reliability statement lands on exactly this large-effect group - 'Some supplements with a limited number of studies and participants suggested large treatment effects' - and 'The Grading of Recommendations Assessment, Development and Evaluation suggested a wide range of quality evidence from very low to high', with the overall pooled conclusion rated very low. Abstract-grade (Br J Sports Med is not in the PMC open-access subset). |
| Wai HS et al 2025 · BMC Complement Med Ther | meta-analysis | supports | moderate | Network + pairwise meta-analysis of RCTs of turmeric products in knee OA; 17 RCTs were included overall, but the WOMAC-pain analysis rests on 7 studies (N=913), of which 5 used bioavailability-enhanced curcuminoid preparations (BE) and 2 conventional (CT). BE vs placebo MD -2.47 (95%CI -3.27 to -1.67) network, -2.47 (-3.25 to -1.68) pairwise = 29.7% change from baseline, exceeding MCID; BE is the only preparation with direct head-to-head placebo data, so the CT and CT+AC estimates vs placebo are indirect. Authors rate direct estimates moderate certainty and network estimates low/very low (imprecision + incoherence); sensitivity analyses excluding high-risk-of-bias and small studies were consistent. |
| Daily JW, Yang M, Park S 2016 · J Med Food | meta-analysis | supports | low | SR+MA of 8 RCTs of turmeric/curcumin for joint arthritis (7 OA, 1 rheumatoid). Placebo-controlled PVAS pool is 3 of the 8 trials, 60 curcuma vs 62 placebo: MD -2.04 (95% CI -2.85, -1.24). Those 3 include the rheumatoid trial (Chandran 2012) and a trial of a turmeric polysaccharide extract 'that contained no curcumin' (Madhu 2013), so the osteoarthritis-plus-curcuminoid placebo signal is essentially Panahi 2014 (curcumin n=21 vs placebo n=19). Second placebo pool (308 vs 291) reports MD -15.26 (-26.94, -3.77) but pools PVAS (0-10) with WOMAC pain (0-100) as a raw mean difference, the authors noting 'their scales were different'; the abstract reports the same analysis as four studies and -15.36 (-26.9, -3.77). No knee-OA-only pooled estimate is computed. Quality low: authors' own verdict is that 'the total number of RCTs included in the analysis, the total sample size, and the methodological quality of the primary studies were not sufficient to draw definitive conclusions'; 'no publication bias' is asserted from an eyeballed symmetric funnel of 3-5 points while the discussion says 'there is still some possibility of publication bias'; first author is President of Daily Manufacturing, Inc., a dietary-supplement manufacturer. |
| Zeng L, Yang T, Yang K, Yu G, Li J, Xiang W, Chen H 2022 · Front Immunol | meta-analysis | supports | low | OA-SPECIFIC, vs-PLACEBO subgroups (Table 2) - the only comparison that bears on this claim: VAS MD -1.33 (95% CI -2.23 to -0.43, P=0.004, I2=94%) from 6 studies / 431 participants; WOMAC-pain -0.66 (-0.88 to -0.43, P<0.0001, I2=34%) from 4 studies / 315; WOMAC-physical-function -0.79 (-1.27 to -0.31, I2=75%) and WOMAC-stiffness -0.35 (-0.57 to -0.12, I2=26%), 4 studies / 315 each. DENOMINATOR: the abstract's headline 'Twenty-nine (29) RCTs involving 2396 participants and 5 types of arthritis were included' spans RA, OA, AS, JIA and gout - no OA pain pool here exceeds 431 participants. The narrative 'summary' SMDs (VAS -2.03; WOMAC-pain -0.69; function -1.65; stiffness -0.22) pool placebo-, NSAID- and add-on-NSAID-controlled trials into one figure and are NOT placebo-controlled effects; versus NSAIDs alone the effect is null (VAS -0.07 [-0.32, 0.19] P=0.62, 2 studies/230; WOMAC-pain 0.04 [-0.18, 0.25], 1 study/331) - a small non-significant subset, i.e. weak evidence of equivalence rather than demonstrated equivalence. ROLE = duplicate-population, ZERO WEIGHT: on the OA arm this paper re-pools the SAME TEAM'S OWN osteoarthritis meta-analysis, s34017975 (Zeng L, Yu G, Yang K, Chen H; Biosci Rep 2021; already group_id zeng-oa), over the same primary trials - Belcaro, Haroyan, Hashemzadeh, Heidari, Henrotin, Kuptniratsaikul, Madhu, Nakagawa, Panahi, Rahimnia, Shep and Srivastava appear in both - and reports four estimates bit-for-bit identical to it: WOMAC-pain vs placebo SMD -0.66 (-0.88 to -0.43) with the same I2=34%, P=0.21; WOMAC-pain vs NSAIDs 0.04 (-0.18 to 0.25); adverse events vs placebo RR 1.18 (0.71-1.94); adverse events vs NSAIDs RR 0.55 (0.34-0.88). An identical estimate with identical heterogeneity is the same analysis, not a replication, so counting this row alongside s34017975 is a literal double count of one team's single pass over one literature. Its genuinely new pooling (RA DAS28/ESR/CRP/RF, AS, JIA, gout) is off this claim's object. QUALITY low, not moderate: unlike s34017975 - same team, weaker instrument the second time - this paper runs no publication-bias test, no GRADE and no sensitivity analysis anywhere in its full text; the placebo VAS pool carries I2=94%; the authors' own conclusion is 'However, due to the low quality and small quantity of RCTs, the conclusions need to be interpreted carefully'; and their limitation (3) states 'The RCTs included in this study were at high risk of bias. The authors of some RCTs were funded by drug manufacturers or were employees, which may introduce bias.' Protocol CRD42022286421; no commercial COI declared. Direction is unchanged and real - every placebo-controlled OA endpoint favours curcumin - this row simply must not vote twice for it. |
| Jhun JY, Lee D, Na HS, Cho KH, Lee SY, Lee JS, Lee YJ, Kim SJ, Park SH, Cho ML 2025 · J Inflamm (Lond) | animal | supports | low | Rat MIA knee-OA model, 4 arms of n=6 (vehicle, curcumin 100 mg/kg, curcumin+omega-3 110 mg/kg, celecoxib 30 mg/kg), oral days 3-21. The CURCUMIN-ALONE arm is what bears on this claim: 'The PWT and PWL measurements of the curcumin group, the celecoxib group, and the curcumin + omega-3 group were compared with those of the vehicle group, and an anti-nociceptive effect was confirmed', and the curcumin group also showed greater weight bearing than vehicle; OARSI and total Mankin scores were likewise lower with curcumin alone than vehicle. The combination arm outperformed both curcumin alone and celecoxib, so the headline effect is the combination's, not curcumin's. Denominator for every figure is 6 rats per arm. No effect sizes are given anywhere in the text - all results are figure-only, reported as significance stars. |
| Balcı EC et al 2026 · study_type: animal | animal | tested-null | low | Rat distal femoral osteochondral defect model; oral micellar curcumin (250 mg/kg/day) improved macroscopic/histological repair scores and cartilage markers (SOX9, collagen II, aggrecan) vs control at 90 days; no pain outcome directly measur |
| Patel M, Betanzos G, Troka M, Modi J, Nageeb G, Kaye AD, Abd-Elsayed A 2026 · Nutrients | observational | supports | low | Narrative review, no data of its own ('No new data were created or analyzed in this study'), and the authors state that 'Given the narrative nature of the review, no formal systematic selection or risk-of-bias assessment was performed.' On this claim's object it relays rather than measures: 'Curcumin and turmeric extracts have relatively consistently shown improvements in knee pain and function, comparable to those of low-dose NSAIDs', seen on VAS pain and WOMAC. The review then tempers it — nutraceuticals such as curcumin 'may reach MCID in select trials or subgroups, but average effects across studies are modest and frequently fall below clinically meaningful thresholds'. CAUTION on the one number quoted near curcumin: WOMAC pain SMD -0.36 comes from a 2020 meta-analysis of 42 RCTs of supplements POOLED across glucosamine, curcumin and collagen — it is not a curcumin-specific estimate. Zero weight (evidence_role: review). |
| Shelton T, Croner L, Alluis A, Aldrich L 2025 · Integr Med (Encinitas) | observational | supports | moderate | Scoping narrative review of curcumin/turmeric across pain-medicine conditions; concludes most clinical research is in knee OA and reports 'efficacy of turmeric in pain control of knee osteoarthritis,' but calls for more research in other pa |
| Wang S et al 2026 · J Control Release | animal | supports | low | Abstract-grade and permanently so (Europe PMC: isOpenAccess=N, inEPMC=N, no PMCID; J Control Release, subscription). ROLE = off-scope on the SUBJECT, ZERO WEIGHT (rubric axis 1, pooled exposure): the tested entity is a composite in which magnesium is a co-therapeutic, not an excipient - 'we engineered porous PLGA microspheres (Cur-Mg/PLGA MPs) incorporating curcumin-modified magnesium hydroxide nanosheets to effectively neutralize excessive hydrogen ions and sustained the release of Mg-Cur nanocomplexes through microporous structures', with the rationale stated as 'Pathological acidosis and magnesium ion (Mg2+) deficiency have been mechanistically associated with cartilage matrix degradation, abnormal subchondral bone remodeling, and chronic inflammation, collectively driving OA progression.' No curcumin-only comparison is reported in the available text (whether the full paper contains a free-curcumin arm is cannot-tell), so this identical row would vote the same way for a claim that MAGNESIUM reduces OA pain - it measured neither member of the pool separately. THE OBJECT IS MEASURED, and the direction is real for the composite: 'Moreover, in vivo evaluations demonstrated that this microsphere system can significantly alleviate OA-associated pain, effectively suppress osteoclast activation, and substantially maintain cartilage matrix homeostasis'; 'Our in vitro experiments revealed that Cur-Mg/PLGA MPs exhibit remarkable chondroprotective effects under both inflammatory and acidic microenvironmental conditions'; mechanism 'Cur-Mg/PLGA MPs inhibit macrophage-to-osteoclast differentiation through ACOD1-mediated mitochondrial metabolic rewiring'. DENOMINATOR/ENDPOINT: the abstract reports no group sizes, no rat strain, no OA induction model, no administration route and no numeric effect estimate for any outcome - the only quantitative anchor is 'significant efficacy in rat models'; which nociceptive test produced 'OA-associated pain' is cannot-tell. Grade stays animal (in vitro + in vivo rat) and quality stays low. Metadata read off the paper: J Control Release 2026 Feb;390:114523, doi 10.1016/j.jconrel.2025.114523, epub 2025-12-13. |
| Dai N et al 2026 · study_type: mechanism | animal | supports | low | Mouse OA model + in-vitro chondrocytes; curcumin elevated miR-338-3p, suppressed EIF4A1, reduced chondrocyte apoptosis/inflammatory cytokines (IL-6, TNF-a, IL-1b) and improved OARSI scores. |
| Shi C et al 2026 · study_type: meta-analysis | meta-analysis | supports | moderate | Umbrella review of 10 meta-analyses/SRs on curcumin formulations for OA; found significant improvements in VAS pain and WOMAC function, efficacy signals comparable to NSAIDs with better tolerability, but only 3/10 rated high quality on AMST |
| Zeng L, Yu G, Hao W, Yang K, Chen H 2021 · Biosci Rep | meta-analysis | supports | moderate | SR + pairwise meta-analysis, 15 RCTs / 1621 participants, searched to Oct 2020 (4 Chinese + 5 English databases). Placebo-controlled subgroups favour curcumin/Curcuma longa extract: WOMAC score-pain SMD -0.66 (95% CI -0.88 to -0.43, P<0.00001, I2=34%) within a 6-study / 806-participant pool, and VAS WMD -11.55 (95% CI -14.3 to -9.06, I2=69%) within a pool Table 4 labels 823 participants / 10 studies. Versus NSAIDs the comparator subgroup shows no difference (VAS WMD -0.34, 95% CI -1.25 to 0.57, P=0.46, I2=0%; WOMAC-pain SMD 0.04, -0.18 to 0.25) - a non-significant small subset, i.e. weak evidence of equivalence rather than demonstrated equivalence. CORRECTION: the previous extract's 'fewer GI adverse events' is not in this paper; it pooled ALL-CAUSE adverse events, and the overall summary is NULL (RR 0.77, 95% CI 0.56-1.05, P=0.10, 1410 participants / 14 studies), with the reduction confined to the NSAID-comparator subgroup (RR 0.55, 0.34-0.88) and nothing versus placebo (RR 1.18, 0.71-1.94). Quality moderate not high: authors' own GRADE is 'high to moderate', downgrading VAS, WOMAC-function and adverse events with the footnote 'Downgraded one level due to the probably substantial heterogeneity'; limitation (8) concedes 'the heterogeneity of some outcomes are high'; and the authors state their higher GRADE is 'possibly because our assessment of the risk of bias is lower' than prior reviews of the same trials. Offsetting positives: Egger/Harbord tests run on all five primary outcomes and all null (VAS P=0.125, WOMAC-pain P=0.301), Cochrane RoB tool, PRISMA, no funding and no declared competing interests. Note also that every 'summary' estimate pools placebo-, NSAID- and add-on-controlled trials together, so the GRADE-high headline SMD -0.57 is not a placebo-controlled effect; the placebo subgroup -0.66 is the interpretable figure for this claim. No protocol registration is stated in the text (not-stated, not absent). |
| Zhou Y et al 2025 · Gels | animal | supports | low | Full-text re-read 2026-08-25 (PMC12840833, Gels 2026;12(1):7). ROLE = off-scope on the OBJECT, ZERO WEIGHT, matching the sibling ruling already applied to s41303695 on this claim: the object is osteoarthritis PAIN, and this paper reports no analgesic outcome. Its in-vivo endpoints are Pelletier synovitis / OARSI / Mankin histology scores, Collagen II and MMP13 immunohistochemistry, micro-CT BV/TV and joint-space width, and a single behavioural test that the paper itself presents as MOTOR function: 'Behavioral evaluation of motor function based on “latency in thermal withdrawal test” response (in seconds).' The apparatus IS nociceptive - 'the soles of the left hind limbs were stimulated with a thermal pain detector' and 'The foot retraction time was also recorded' - but the reported direction cannot be read as analgesia: 'ACLT surgery resulted in significant motor impairment, as indicated by the increased “latency in thermal withdrawal test” values. All treatments improved performance, with Cur@HBC resulting in the most significant recovery.' OA therefore LENGTHENED withdrawal latency and curcumin SHORTENED it, which on the standard nociceptive reading is reduced thermal sensitivity in the OA animals and restored sensitivity after treatment - the opposite of a pain benefit. The authors read it as slower movement instead ('Cur@HBC significantly improved joint function, as evidenced by a decreased response time in motor testing') and claim analgesia nowhere in the paper; the word pain appears only in the Introduction's disease description and in mechanistic background about M1 macrophages and 'pain sensitization'. Either reading leaves this claim's object unmeasured, which is why this is off-scope rather than contradicts. CORRECTION to the previous extract: the model is MICE, not rats - 'Thirty-six C57BL/6 (male, 4 weeks old) mice were purchased from the Guangdong Experimental Animal Center (Guangzhou, China). Six animals were randomly assigned to each group.' DENOMINATORS: six mice allocated per group, but every histology and IHC figure reports 'Data are presented as the mean ± SD (n = 5)', the macrophage flow-cytometry figure reports n = 3, and Figure 4 - the micro-CT panels and the behavioural latency - states no n at all; 36 mice at six per group implies six groups while only five ever appear in the figures. No numeric effect estimate exists for any in-vivo endpoint; all are figure-only significance stars. SUBJECT is not pooled - a free-curcumin arm exists ('Two weeks post-surgery, a total volume of 10 μL of saline, HBC, curcumin (20 μg/mL), or Cur@HBC was injected into the knee joint cavity via a microinjector') - but the headline effect belongs to the composite and the vehicle is not inert: 'Interestingly, the HBC carrier without curcumin also exerted protective effects, suggesting that the carrier itself may have cartilage-supportive or lubrication function.' Note also that the exposure is intra-articular injection in a mouse knee, not the oral standardized extract this claim's scope note describes. QUALITY stays low, now on stated grounds rather than by default: no blinding or randomisation of outcome assessment is described anywhere in the full text (the word 'blind' does not occur), no effect sizes, an unexplained six-allocated-versus-five-analysed gap, and 30 days from submission to acceptance in an MDPI journal. Funding NSFC 32503103 / Guangdong 2025A1515012160 / Guizhou gzwkj2025-617; 'The authors declare no conflicts of interest'; ChatGPT 5.2 disclosed for language polishing. DIRECTION on what the paper DID measure is real and one-way - structure, matrix, cytokines, chondrocyte apoptosis and M2 polarisation all favour curcumin - so the stance is left as filed; it is the weight, not the sign, that this re-read removes. |
| Park HS et al 2026 · study_type: animal | animal | supports | moderate | Rat MIA-induced OA model; curcuminoids phospholipid (CP) improved pain-related weight-bearing behavior and cartilage integrity vs MIA control, comparable direction to celecoxib positive control. |
| Hridayanka KSN et al 2025 · Int J Mol Sci | animal | supports | low | Rat MIA-induced knee OA (2 mg monoiodoacetate, left knee), 9 male SD rats randomised 3 per group — saline control, MIA+vehicle, MIA+intra-articular curcumin nanoemulsion 30 mg/mL weekly x4 wk. NO PAIN OR NOCICEPTIVE ENDPOINT IS MEASURED; every result is structural, inflammatory or in-vitro, and every in-vivo figure below has n = 3/group as its denominator: 'In the MIA model, localized delivery of n-Cur significantly reduced knee joint edema and joint space narrowing in the target site'; Kellgren-Lawrence grade 'MIA-induced vs. Cur nanoemulsion: 3.33 ± 0.21 vs. 1.33 ± 0.21 (a.u.)'; modified Mankin score '9.70 ± 0.40 vs. 4.05 ± 0.66' (scored 'by three independent blinded observers'); sub-synovial infiltration '3.75 ± 0.25 vs. 2.25 ± 0.25 a.u.'; serum MDA '0.92 ± 0.04 vs. 0.59 ± 0.03 a.u.'; and in vitro 'Both curcumin and nanoemulsion reduced macrophage migration despite the presence of IL1β'. All results point one way — nothing in the paper opposes curcumin, so the prior 'mixed' stance had no basis in the paper. The only pain sentences are Discussion inference citing refs 43-45. Authors' own limitations: 'Our study is limited to the fact that animal numbers in this study were significantly reduced' and 'A comparative assessment with native curcumin in vivo and further analysis of cartilage histology could not be achieved'. Read from PMC full text (PMC12653435). |
| Tuntiyatorn P et al 2025 · BMC Complement Med Ther | RCT | tested-null | low | Pilot double-blind placebo-controlled RCT in hand OA, n=27 randomised (13 curcumin / 14 placebo, 88% female, age >50), 3 months of whole turmeric powder capsules ('equivalent to 28.75 mg of curcuminoids per capsule', 'administered as two capsules, three times daily for three months' = ~172.5 mg curcuminoids/day) plus a 3-month washout. THE RANDOMISED COMPARISON IS NULL: 'no significant differences were observed in VAS hand pain at rest, VAS hand pain during activity, VAS hand morning stiffness, DASH score, hand grip, pinch strength and rescue drug usage at 1, 3, and 6 months'. The significant figures previously stored in this extract (rest 1.37±0.64 p=0.032; activity 1.74±0.73 p=0.017; DASH 11.84±5.88 p=0.044) are NOT treatment effects - they come from Table 3, 'The mean differences within-curcumin group at 3 months compared with baseline', whose denominator is the 13-patient curcumin arm alone (12 completing, after one withdrawal for severe dyspepsia), with no placebo contrast. The Results narrative's 'compared to the placebo group' contradicts its own table title, and the abstract Conclusion repeats the error. Stance stays tested-null rather than contradicts because the trial disclaims power for this comparison: 'As this was a feasibility study, no conventional sample size calculation was performed, and the study was not powered to test the relative efficacy of curcumin versus placebo.' After discontinuation 'both VAS and DASH scores returned to baseline by the 6-month mark'. Safety labs (CBC, creatinine, LFTs, ESR, CRP) unchanged in both arms. |
| Park D et al 2026 · study_type: observational | observational | mixed | moderate | Structured narrative review (no meta-analysis) of nutraceuticals incl. curcumin for knee OA with sarcopenia; symptomatic benefits across curcumin/glucosamine/Boswellia/omega-3/collagen described as 'modest, heterogeneous, and formulation-de |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.