Curcumin / turmeric extract

Turmeric is about 3 percent curcuminoids by weight. Curcumin is the largest of the three, and it carries the largest randomized base of any compound in this guide — roughly forty knee-osteoarthritis trials, more than a dozen meta-analyses, and in the largest network analysis, 2,175 patients.

It is also one of the worst-absorbed compounds anyone has ever tried to sell as a supplement. Those two facts are the whole entry: there is a lot of evidence, and most of it was generated at exposures nobody can specify, using products that are not interchangeable.

What curcumin does to chondrocytes

The mechanism is clean and well characterised. Curcumin inhibits NF-κB, the transcription factor that carries most of the catabolic signalling in an inflamed joint. In IL-1β-stimulated human articular chondrocytes it blocks the upregulation of MMP-3 and the nuclear translocation of NF-κB — and, more interestingly, reverses the IL-1β-driven fall in type II collagen. It protects matrix synthesis rather than merely suppressing breakdown.

Combined with resveratrol the two reach NF-κB by different routes — curcumin by shutting down IκB kinase and leaving the proteasome untouched, resveratrol the other way about — and together suppress COX-2, MMP-3, MMP-9 and VEGF and block caspase-3. Where the pair was actually tested against each compound alone, on type II collagen and Sox-9, it beat both. A curcuminoid extract cuts IL-6, iNOS, COX-2, MMP-3 and the aggrecanases ADAMTS-4 and ADAMTS-5 in bovine and osteoarthritic human chondrocytes — and in that experiment the hydrolysed collagen and green tea extract it was mixed with contributed almost nothing. Injected directly into a rat knee after ligament transection, curcumin improves cartilage histology and suppresses TLR4 and NF-κB in the joint.

The chondrocyte experiments run at 50 micromolar. An independent human pharmacokinetic study measured unconjugated curcumin below 2 nanomolar in most plasma samples — including at the highest dose tested and with black pepper — and the best formulation's brief peak of 6.7 to 38 nanomolar was still about a hundredfold short of the concentrations used in the dish. That is a gap of roughly four orders of magnitude between where the mechanism was demonstrated and where a capsule puts you.

There is also a standing chemical objection: curcumin is classed as a pan-assay interference compound and an "invalid metabolic panacea" — unstable, reactive with many proteins non-specifically, essentially non-bioavailable — which its critics argue makes a large share of the in-vitro literature artefactual. That paper's clinical claim, that no double-blind placebo-controlled trial had succeeded, was contestable when it was published and has since been contradicted by knee trials. The charge against the mechanism literature is separate, and it still stands.

The trials against ibuprofen and diclofenac

This is where curcumin's reputation comes from, and the comparisons are real.

The largest single trial (n=367, Thai government-funded) found turmeric extract 1,500 mg/day met the non-inferiority margin against ibuprofen 1,200 mg/day at four weeks on WOMAC total, pain and function — but not on stiffness, which fell short. The share of patients reporting any adverse event was the same in both arms; abdominal pain and distension were the one event that separated. A second trial (n=139, 28 days) found curcumin 500 mg three times daily and diclofenac level on pain at day 28, with adverse events of 13 percent versus 38 percent and no need for H2-blocker rescue. That trial was powered to detect a difference rather than to rule one out, and set no non-inferiority margin, so the matched pain scores are a test that failed to separate them rather than a demonstration that they are alike. A turmeric-pepper-ginger combination matched naproxen on serum prostaglandin E2 at four weeks — though PGE2 was the only endpoint reported, and there was no placebo arm, so two ineffective arms would look the same.

Pooled, the pattern holds. Curcuminoids show no significant efficacy difference from NSAIDs with significantly fewer gastrointestinal adverse events; turmeric extracts match NSAIDs with 12 percent fewer adverse events overall; and the Bayesian network analysis puts adverse reactions at an odds ratio of 0.51 (95% CI 0.25 to 0.94) against NSAIDs.

The caveat that travels with all of it: the head-to-heads mostly lack a placebo arm, and the two biggest are four weeks and 28 days long.

The placebo-controlled trials

The most rigorous is Wang 2020 (n=70, publicly funded, Annals of Internal Medicine): Curcuma longa extract reduced pain by 9.1 mm on a 100 mm visual analogue scale, 95% CI −17.5 to −0.7 — a confidence interval whose upper bound nearly touches zero — and produced no change in MRI effusion-synovitis volume or in cartilage composition at twelve weeks. That is the only time anyone has imaged the joint in a curcumin trial, and the 16-trial systematic review confirms that no imaging or biochemical outcome separated from control in any included study.

The rest of the placebo-controlled base is broad and mostly positive. Curcuminoids 1,500 mg/day with piperine improved WOMAC, VAS and the Lequesne index at six weeks (n=40). Theracurmin at 180 mg curcumin/day did not lower pain against placebo across everyone analysed at eight weeks, did once the six patients who started near zero were excluded, and cut celecoxib use (n=50). Curcugen 1,000 mg/day improved KOOS pain, the timed up-and-go and the six-minute walk, though not the chair stand or the fast-paced walk (n=101). Nanomicelle curcumin at 80 mg/day improved total WOMAC and every subscale (n=71). Bio-optimised curcumin missed both of its co-primary endpoints on the analysis it had prespecified — patient global assessment and the cartilage-degradation marker Coll2-1 improved in every arm including placebo, with no difference between treatments — and reached placebo only in post-hoc analyses that pooled its two dose arms (n=150). A three-arm trial found curcumin plus boswellic acid beat placebo on WOMAC pain and all four performance tests, while curcumin alone beat placebo on two performance tests and nothing else; the two active arms were never tested against each other (n=201). Meriva phytosome, in the field's longest follow-up at eight months, improved WOMAC, Karnofsky index and treadmill walking distance — in a non-randomised, unblinded design.

Pooled across 16 RCTs and 1,810 patients, turmeric extracts reduce pain with a standardised mean difference of −0.82 (95% CI −1.17 to −0.47) and improve function at −0.75. Restricted to trials of curcuminoids given alone — 15 RCTs, 1,670 patients — pain falls 1.77 points on the VAS and total WOMAC by 10.47. Only those two of its five pooled measures cleared the threshold for a change a patient would notice. Higher BMI predicts less benefit.

Three results cut the other way. Adding curcumin 1,000 mg/day to diclofenac changed nothing: VAS p=0.923, and all five KOOS subscales non-significant. A one-week turmeric trial found no difference between treatment and placebo — the placebo arm improved on the same items. And ATLAS, the 2026 independent Australian trial of a combination containing curcumin 500 mg and piperine 5 mg, found a between-group difference of 0.16 mm on a 100 mm scale (95% CI −6.81 to 7.12) at twelve weeks.

The heterogeneity behind the pooled numbers is I² of 86 percent for pain and 90 percent for function. The NIH-funded independent meta-analysis, having found the same positive direction, concluded that the evidence base is not adequate in size or quality to support any clinical practice recommendation.

Formulation and bioavailability

This is the most practically useful section of the entry, and it is where the labels and the measurements disagree.

One trial has compared the major formulation strategies head to head in people: eight formulations, a single 207 mg curcumin dose, randomized double-blind crossover, twelve healthy adults, university-funded and independent of every product tested. The results, as multiples of the plasma exposure from a plain extract:

• Micelles — 57×. Significant, and the fastest absorbed.
• γ-cyclodextrin complex — 30×. Significant.
• Phytosomes — 7.5×. Not significant.
• Submicron particles — 6.5×. Not significant.
• Adjuvants including piperine — no increase. 18.65 nM·h against 19.06 nM·h for the plain extract.
• Liposomes and turmeric oils — no increase.

No free curcumin was detectable in any subject on any formulation. The mechanistic reading offered by the authors is that what matters is staying soluble after digestion; strategies aimed at blocking metabolism after absorption — which is precisely what piperine does — did not work.

An independent Amsterdam crossover put the specific combination most people buy on the bench: curcumin C3 Complex at 600 mg, at 2,400 mg, and at 2,400 mg with 20 mg of piperine, against a micellar and a solid-lipid product. Free unconjugated curcumin was below the 2-nanomolar detection limit in every participant on all three C3 arms — quadrupling the dose did not change it, and neither did the black pepper. Median total-curcumin exposure was 897 nM·h for C3 Complex at 2,400 mg and 524 nM·h for the same dose with piperine: lower, not higher. Only the micellar product produced measurable free curcumin, and even that peaked about a hundredfold below the concentrations the chondrocyte work uses. The authors' summary of the additive is that it is useless.

Head-to-head against a modern water-dispersible extract, 1,500 mg of curcuminoids with 5 mg piperine taken three times daily gave equivalent total exposure and a third of the peak concentration of a single 250 mg dose.

The piperine practice traces to one 1998 study, which reported a 2000 percent rise in bioavailability from 20 mg piperine alongside 2 g curcumin. It was run by authors affiliated with the company that sells both the extract and the pepper extract, and its serum curves returned to baseline within three hours — a time course no subsequent human study has reproduced.

Other measured multipliers, all against unformulated powder: γ-cyclodextrin 39×, Theracurmin 36–43× with peak concentration at 1.5–3 hours instead of 8, and a phosphatidylcholine complex giving no dose-adjusted plasma advantage but five-fold higher concentrations in gut tissue. Manufacturer trials reporting hundredfold multipliers exist; they are dose-normalised figures computed on conjugated metabolites, which is the measurement problem the independent studies were written to point out.

And then the finding that cuts against all of it. The one network meta-analysis to separate bioavailability-enhanced products from conventional ones found no clinical advantage on pooled WOMAC pain: −3.17 points for conventional preparations, −2.47 for enhanced ones. Blood levels differ by 57-fold; measured pain relief does not follow. Nobody has run the clinical trial that would explain why.

Dose and practical notes

Because the milligram is not the exposure, trialled regimens are best read by formulation:

• Plain standardised extract: 1,500 mg/day of turmeric extract, or 500 mg of curcumin three times daily, with or without 5–15 mg of piperine.
• Absorption-enhanced products, at far lower curcumin mass: Theracurmin 180 mg/day, nanomicelle or SinaCurcumin 80 mg/day, bio-optimised extract at 187 or 280 mg, CurQfen 400 mg/day, Curene 500 mg/day, Curcugen 1,000 mg/day.
• The meta-analytic reference dose is about 1,000 mg/day, and the pooled comparison of 1,000 mg/day and above against anything below it found no difference in pain relief or adverse events.

Benefit appears within four weeks in most trials. The longest osteoarthritis follow-up is eight months; the longest randomized exposure of any kind is 52 weeks, in rheumatoid arthritis. Turmeric powder from the spice rack is roughly 3 percent curcuminoids and is not what any of these trials tested.

Safety

Gastrointestinal tolerability beats NSAIDs consistently and by a measured margin — 13 percent versus 38 percent adverse events against diclofenac, 12 percent fewer than NSAIDs pooled, and rates no different from placebo. Fifty-two weeks of curcumin 1 g with piperine 5 mg twice daily produced no serious adverse effects.

Liver. Turmeric supplements are an established cause of rare, idiosyncratic drug-induced liver injury. The US Drug-Induced Liver Injury Network adjudicated ten cases, all enrolled since 2011 and six since 2017: hepatocellular injury in nine, five hospitalised, one death from acute liver failure, with a latency of one to four months. Seven of the ten carried the HLA-B*35:01 allele, at a frequency of 0.450 against 0.056–0.069 in population controls — strong evidence that this is an immune-mediated reaction in genetically susceptible people rather than dose-dependent toxicity. Three of the seven products chemically analysed also contained piperine, and seven Italian pharmacovigilance cases all involved high-dose, high-bioavailability formulations. Recent case reports continue to name turmeric-plus-black-pepper products specifically.

The size of that signal matters as much as its existence. Numerous clinical trials report no organ toxicity or serious adverse events, published cases usually involve concomitant medications, and injury generally resolves when the supplement is stopped. Against roughly 15.6 million US adults taking one of the six hepatotoxicity-associated botanicals in any given month — turmeric the most common, and arthritis the reason its users reach for it — ten adjudicated cases over eighteen years is a rare event. The US Pharmacopeia nonetheless added a cautionary statement to its turmeric and curcuminoid monographs in 2026: consult a clinician before use if you have a history of liver problems, and stop and seek advice on abdominal pain, dark urine or jaundice.

Piperine and drug interactions. Piperine inhibits P-glycoprotein and CYP3A4 in vitro, the two systems responsible for much of the first-pass clearance of oral drugs. The only human test of that concern was negative: 4 g of curcuminoids with 24 mg of piperine over two days produced no meaningful change in the pharmacokinetics of midazolam, flurbiprofen or paracetamol, and no change in midazolam's sedative effect. The unresolved half is the opposite direction — piperine also activates the pregnane X receptor and induces CYP3A4 and MDR1 in human hepatocytes and intestinal cells, and whether chronic supplemental piperine does that in people has not been measured. Where this would matter is narrow-therapeutic-index drugs: tacrolimus, ciclosporin, digoxin, warfarin. None has been tested with piperine in humans.

Other cautions carried from the trial literature and product labelling: an antiplatelet effect, so care with anticoagulants and before surgery; gallbladder contraction, so care with gallstones; and avoidance of supplemental doses in pregnancy.

Curcumin outside the joint

People take curcumin for more than knees, and the entry's cartilage tier says nothing about the rest of it. Pooled across 66 randomized trials, curcumin lowers CRP by 0.58 mg/L, TNF-α by 3.48 pg/mL and IL-6 by 1.31 pg/mL, raises total antioxidant capacity and lowers malondialdehyde. In metabolic syndrome, 1,000 mg/day of C3 Complex curcuminoids with black-pepper extract at 100:1 lowered LDL-C, non-HDL-C, total cholesterol, triglycerides and Lp(a) and raised HDL-C over eight weeks. Six trials pool a modest antidepressant effect in major depressive disorder.

Two results complicate the anti-inflammatory story from inside it. A 19-trial meta-analysis across rheumatic, renal, metabolic and cardiovascular disease found no significant reduction in CRP, hsCRP, IL-1β, IL-6 or TNF-α. And in the one osteoarthritis cohort where symptoms and markers were measured together, no inflammatory marker differed between the curcuminoid and placebo arms — the authors state plainly that the symptom improvement seen in those same patients cannot be attributed to systemic anti-inflammatory action. In rheumatoid arthritis, the largest and longest curcumin-plus-piperine trial ever run (n=200, 52 weeks) found flare-free survival of 60 percent against 64 percent on placebo while DMARDs were tapered, despite confirming that adequate serum curcuminoid levels were reached.

What would move this entry

Two experiments would change the tier, and both are obvious enough that their absence is the story.

The first is a structural trial. One twelve-week MRI study is the entire structural literature, and it was null on effusion-synovitis and cartilage composition. A longer trial on an absorption-enhanced formulation, with cartilage volume or composition as a primary endpoint, would settle whether the chondrocyte biology survives the journey.

The second is a formulation head-to-head with a clinical endpoint. The pharmacokinetics say exposure differs by 57-fold between products; the pooled clinical data say pain relief does not follow. One of those two observations is misleading, and nobody has run the trial that would say which.

What is established meanwhile is worth stating plainly: for symptomatic knee osteoarthritis over four to sixteen weeks, curcumin relieves pain with no measurable disadvantage against ibuprofen or diclofenac, and with markedly fewer gastrointestinal events, across a randomized base larger than any other supplement in this guide.