Turmeric (curcumin)

What turmeric is

Turmeric, also called yellow root, is the rhizome of Curcuma longa. According to the German Federal Institute for Risk Assessment (BfR), it contains about 1 to 5 percent curcuminoids, of which curcumin makes up around 75 to 80 percent. Purified curcumin is approved as the food color E100 and must contain at least 90 percent curcuminoids for this purpose.

Spice and extract are orders of magnitude apart: the European pharmacopoeia authority EDQM found an average of 2.34 percent curcuminoids in 21 batches of the root. A capsule is something different from golden milk.

The bioavailability problem

Curcumin is poorly absorbed and quickly converted in the gut and liver. How poorly is shown by a dose escalation in healthy volunteers: at single doses of 500 to 8,000 mg, no curcumin at all was detectable in serum; only at 12,000 mg were low levels found.

The best known is piperine from black pepper. The famous 2,000 percent come from a 1998 study in which volunteers received a single dose of 2 g of curcumin with 20 mg of piperine. In the animal experiment of the same study it was 154 percent, in another rat experiment a factor of 1.2 to 1.7, and in a third there was no effect. The BfR therefore states that the increase varies greatly depending on dose and dose ratio.

The technical formulations have been measured more cleanly. In a crossover in 13 women and 10 men, each given 500 mg of curcuminoids, micellar curcumin reached 185 times the area under the concentration curve of native curcumin, and a lecithin formulation 29 times the total absorption. Better efficacy does not follow from this: in the lecithin study, concentrations stayed below what would be needed to inhibit most of the targets studied, and an umbrella review of 10 reviews rules out the comparison between formulations because direct comparative studies are lacking.

How it is supposed to work

Curcumin inhibits NF-kappa-B and thereby the formation of pro-inflammatory messengers. In cell culture and animal models, this holds true. The dispute is about whether the concentrations required for this occur in humans after a capsule. A widely cited 2017 paper classified curcumin as a non-specifically reacting screening molecule and stated that despite more than 120 clinical trials, no double-blind trial had been successful. The later osteoarthritis meta-analyses have overtaken this statement, but not the gap between the dish and the knee joint.

In humans, the inflammatory markers do at least move: a meta-analysis of 21 studies with 1,705 knee osteoarthritis patients found a standardized mean difference of -0.906 for CRP and of -0.921 for TNF-alpha, but no difference for interleukin-6 and prostaglandin E2.

What the liver has to do with it

Among the promised benefits, support for liver detoxification comes up regularly. No study tests a detoxification endpoint, and it does not exist as a measurable quantity. What is available on the liver concerns fatty liver: a network meta-analysis of 27 studies with 1,691 participants found better regression of steatosis (odds ratio 4.39), and a second one of 15 studies with 835 participants found no effect at all on ALT.

At the same time, the liver is the only organ with a clear harm signal: LiverTox, the reference database of the US National Institutes of Health, lists turmeric in the highest category, A. A substance that can lower liver values can also damage the liver in some individuals.

What the studies show

Network meta-analysis on knee osteoarthritis, 2024

Bayesian network meta-analysis of 23 randomized trials from 7 countries with 2,175 patients. Compared with placebo, curcumin reduced pain on the visual analog scale by 1.63 points (95% confidence interval -2.91 to -0.45) and the WOMAC total score by 18.85 points (-29.53 to -8.76). All endpoints are patient-reported scores, not the joint findings.

Curcuma extract versus ibuprofen, 367 patients

Multicenter randomized non-inferiority trial: 185 participants received 1,500 mg of Curcuma domestica extract per day, 182 participants 1,200 mg of ibuprofen per day, each for 4 weeks. At week 4, the extract was non-inferior for WOMAC total (p = 0.010), pain (p = 0.018) and function (p = 0.010); for stiffness, the threshold was narrowly missed (p = 0.060). Abdominal pain was more frequent on ibuprofen (p = 0.046). There was no placebo group.

Ten cases of liver injury from DILIN

Case series of all cases adjudicated between 2004 and 2022 in the US Drug-Induced Liver Injury Network with turmeric as the implicated product: 10 cases, all enrolled since 2011, 8 of them women, median age 56. In 9 of 10, the injury was hepatocellular, 5 patients required hospitalization, 1 patient died of acute liver failure. Turmeric was confirmed in all 7 products tested, 3 of which also contained piperine. 7 of the 10 patients carried HLA-B*35:01, allele frequency 0.450 versus 0.056 to 0.069 in controls.

Where the data stop

The osteoarthritis evidence is broad but thinly built. A critical appraisal of 7 systematic reviews assessed 48 endpoints and rated 37 of them as very low quality of evidence; in a meta-analysis of 11 studies with 1,009 participants, most studies had fewer than 100 participants. A dose-response relationship is also missing: in 1,258 participants, high and low curcuminoid doses did not differ.

For mood, the effect is present but shaky: the largest meta-analysis of 19 studies found a standardized mean difference of -0.76 for depression with a heterogeneity of 94.5 percent and itself calls the finding fragile; an earlier analysis of 10 studies arrived at -0.32 and rated the quality of evidence as low.

Brain protection holds up only for memory: in 18 double-blind studies, the standardized mean difference there was 0.57; for executive function and attention, nothing emerged. The longest controlled trial is the most sobering: 88 patients, 12 months, 2,000 mg per day of a bioavailability-optimized formulation, and no difference from placebo either in vascular function (p = 0.69) or in cognition.

There are no figures on liver detoxification, because the endpoint does not exist. On cancer prevention and anti-aging, there is nothing comparable to the osteoarthritis finding.

Status, approval and legal

In the EU, Curcuma longa has a monograph as a traditional herbal medicinal product, with the indication relief of digestive complaints such as feelings of fullness, sluggish digestion and flatulence. The EMA’s HMPC committee explicitly states that the data are not sufficient for the higher classification as well-established use; the maximum daily dose of 4 g of turmeric corresponds to at most 209 mg of curcuminoids. For intake from all sources, an ADI of 3 mg per kilogram of body weight per day applies. That it is exceeded in practice is documented: the BfR assessed two products that, at 70 kg body weight, came to 8.3 and 6.1 mg of curcumin per kilogram per day. There is no authorized health claim, because the European Commission has suspended the assessment of botanical claims since September 2010. Neither the EMA nor the BfR nor the Danish assessment comments on doping.

Safety

The central issue is the liver. LiverTox lists turmeric with likelihood score A, the highest category, as a well-documented cause of clinically apparent liver injury; latency is typically 1 to 4 months, the pattern is hepatocellular with transaminases often above 1,000 U/L, and HLA-B*35:01 is found in more than 70 percent of cases versus 10 to 15 percent in controls. From Italy, 7 suspected cases from Tuscany are documented, 4 of them with piperine-containing products, plus 18 cases from the Italian phytovigilance system. The EMA classifies this as follows: in all cases, other products were taken at the same time, a causal relationship is not established, but the data suggest that hepatotoxicity is a risk of highly bioavailable and high-dose preparations. Because of its effect on bile secretion, turmeric is not recommended in biliary obstruction, cholangitis, liver disease, gallstones and other biliary diseases; it is advised against in pregnancy and breastfeeding, and for people under 18 its use is not established due to lack of data. Possible interactions exist with NSAIDs, platelet aggregation inhibitors, lipid-lowering drugs and immunosuppressants and, according to one case report, with warfarin; the EMA considers the clinical evidence for this insufficient.

Sources

• Zhao et al., J Ethnopharmacol 2024 (network meta-analysis, knee osteoarthritis)
• Kuptniratsaikul et al., Clin Interv Aging 2014 (Curcuma extract vs. ibuprofen)
• Chen et al., Front Pharmacol 2025 (critical appraisal of the reviews)
• Halegoua-DeMarzio et al., Am J Med 2023 (DILIN case series on liver injury)
• LiverTox, NIDDK/NIH: Turmeric
• Shoba et al., Planta Med 1998 (piperine and bioavailability)
• Schiborr et al., Mol Nutr Food Res 2014 (micellar curcumin)
• Cuomo et al., J Nat Prod 2011 (lecithin formulation)
• BfR, opinion No. 040/2021 of December 14, 2021: curcumin in food supplements
• EMA/HMPC, addendum to the assessment report Curcuma longa L., rhizoma, November 19, 2025