Pomegranate

Most foods in this section have a mechanism and no trials, or trials and no mechanism. Pomegranate has both, at small scale, and it also has the clearest example on the site of a problem that quietly undermines the whole polyphenol literature: the compound that works in the dish is not the compound in the food.

The chondrocyte pathway

Pomegranate's polyphenols are hydrolyzable tannins — punicalagins and ellagitannins. In primary chondrocytes taken from human osteoarthritic cartilage, a standardized extract blocked the MKK3 → p38α-MAPK → RUNX-2 route, which is the signal that turns a chondrocyte hypertrophic and matrix-degrading. A second study from the same laboratory showed suppression of IL-6 and reactive oxygen species by blocking IKKβ, IκBα degradation and NF-κB/p65 translocation — and NF-κB-inducing kinase alongside them, so both arms of that pathway. Punicalin keeps FOXO3 in the nucleus of cytokine-stimulated chondrocytes.

The whole-animal work is oral, which matters. Rabbits with surgically transected cruciate ligaments, fed pomegranate extract in their drinking water for eight weeks, lost less Safranin-O staining, showed less chondrocyte clustering, had lower MMP-3, MMP-9 and MMP-13 in cartilage and lower IL-6, MMP-13 and PGE2 in synovial fluid, and expressed more aggrecan and type II collagen. In mice given the juice itself rather than an extract, chondrocyte damage and proteoglycan loss fell in a dose-dependent way — at 20 mL/kg, which scales to well over a litre of juice a day in a person.

The trials

Thirty-eight people with knee osteoarthritis drank pomegranate juice for six weeks against a control group that drank nothing. WOMAC total, stiffness and physical function all improved within the juice group. Two measures were reported against the control and both separated: serum MMP-13 fell (p = 0.02) and glutathione peroxidase rose (p = 0.02). Without a placebo beverage the symptom scores are unblinded, so MMP-13 — an enzyme, not an opinion — is the part of that trial that carries weight.

A properly double-blind trial exists but tests something else. Sixty obese women with knee OA took 500 mg of pomegranate peel extract twice daily for eight weeks alongside their usual drugs. Against placebo, hs-CRP (p = 0.002), NF-κB (p = 0.045), MMP-1 (p = 0.040) and MCP-1 (p = 0.003) all fell. Every endpoint was a blood marker; no pain, function or imaging outcome was reported. And peel extract in a capsule is not pomegranate — the peel carries most of the punicalagin and nobody eats it.

Pooled, a 2026 systematic review of seven studies found four randomized OA trials improving WOMAC, KOOS and VAS with lower IL-6 and TNF-α, no serious adverse events, and an overall certainty of evidence graded low to very low on small samples, heterogeneous interventions and short follow-up. An earlier review of 23 human, animal and cell studies came to the same directional conclusion and the same call for larger trials.

Urolithin A, and why two people get different results

The ellagitannins in pomegranate are poorly absorbed. What reaches the bloodstream in quantity is urolithins — compounds gut bacteria make from those tannins — and urolithin A is what the cell studies keep finding active: in human OA chondrocytes it suppressed NO, PGE2, COX-2, iNOS, TNF-α and IL-6 concentration-dependently, lowered MMP13 and ADAMTS5, and reduced matrix degradation, working through PI3K/Akt/NF-κB. It protected mice with surgically destabilized joints.

Whether you make it depends on which bacteria you carry. People sort into distinct urolithin metabotypes producing quite different metabolites from the same fruit, and the distribution shifts with overweight and metabolic syndrome. That is a real, measured source of person-to-person variation sitting underneath every pomegranate trial ever run, and no trial has stratified by it.

There is at least direct evidence that something bioactive circulates: plasma drawn two hours after an oral dose inhibited IL-1β-induced PGE2 and nitric oxide in chondrocytes and inhibited COX-2 activity, while pre-dose plasma from the same animals did not.

Practical notes

The juice trials use juice; the biomarker trial uses a peel-extract capsule at a dose no fruit delivers. Nobody has compared them head to head. Pomegranate juice inhibits intestinal CYP3A4, which puts it in the same interaction category as grapefruit for drugs metabolised that way — worth checking against a medication list. No serious adverse events were reported in any of the clinical trials.

What would change this entry

Two studies. A pomegranate trial with an imaging or structural endpoint would test whether the MMP-13 signal means anything for cartilage over time. And a trial stratified by urolithin metabotype would test the more interesting hypothesis: if the effect is carried by a metabolite only some people make, the producers should account for the whole result — which would also explain why so many polyphenol trials in this section come back flat.