Green tea

Green tea has one of the largest chondrocyte literatures of any food compound and one of the smallest human ones. Reading the two together is a useful exercise, because the gap between them is not an accident of funding or interest — it is a route-of-administration problem hiding in plain sight.

What EGCG does, and how it was given

In 2002, epigallocatechin-3-gallate was shown to suppress both the activity and the expression of COX-2 and iNOS-2 in human chondrocytes stimulated with IL-1β. Two decades of follow-up work has been consistent, and the best of it measures structure and pain in the same animals: mice with surgically destabilized menisci given EGCG had less Safranin O loss, less cartilage erosion and lower OARSI scores at four and eight weeks, with reduced aggrecan and type II collagen cleavage epitopes, reduced MMP-13 and ADAMTS5 staining, and lower MMP-1, -3, -8, -13, ADAMTS5, IL-1β and TNF-α message — and they behaved as though they hurt less, on von Frey and open-field testing.

The EGCG was injected into the abdominal cavity. The rat study that found less cartilage degeneration through enhanced autophagy injected it into the joint. A further paper on polyphenols and cartilage makes the position explicit in its own title, proposing intra-articular injection as the route to a cartilage therapeutic.

That pattern is the finding. Every delivery method that has produced a cartilage result in this topic goes around the gut, and the gut is where a cup of tea has to work. It is the same gap measured directly for cherry anthocyanins, whose plasma concentrations after drinking are orders of magnitude below what the cell experiments use.

The human trial

Fifty adults with knee osteoarthritis were randomized to green tea extract tablets plus diclofenac, or diclofenac alone, for four weeks. Mean change favoured the green tea arm on VAS pain (p = 0.038), total WOMAC (p = 0.006) and WOMAC physical function (p = 0.004). The WOMAC pain sub-score did not differ (p = 0.163), nor did stiffness (p = 0.150).

Two pain instruments in one trial, pointing different ways. The trial was open-label with no placebo: one group knowingly took an extra tablet, the other knowingly did not, and every endpoint was self-reported. It also tested extract tablets rather than brewed tea, on top of an NSAID, for four weeks. It is the best human evidence in this entry, which is the most useful thing to know about the human evidence in this entry.

The cohorts and the genetics disagree

In 655 people interviewed in Antalya, tea drinkers had less symptomatic knee osteoarthritis (p < 0.05) — one of only two dietary factors in that survey to show anything.

In a two-sample Mendelian randomization across 50,508 people including 10,083 with osteoarthritis, genetically predicted tea intake was associated with higher risk: inverse-variance weighted OR 1.19 (95% CI 1.08–1.30), weighted median OR 1.22 (1.07–1.40), with no directional pleiotropy on MR-Egger and no heterogeneity.

The honest reading is not that tea damages cartilage. A genetic instrument for "tea intake" captures a habit, and in a European cohort that habit is largely black tea taken with milk and sugar, entangled with income, smoking and body weight in ways these instruments do not cleanly separate. What the MR does do is remove the observational association's claim to be causal — which was the only thing it had.

Practical notes

Brewed green tea and green tea extract are different exposures, and the distinction matters for safety as well as dose: concentrated extract carries a liver-injury signal that brewed tea does not, which is why extract products carry warnings in several jurisdictions. This entry does not put a number on that, because no record quantifying it turned up in the search behind it.

What would change this entry

The synovial fluid measurement. For broccoli, someone answered the delivery question directly, by sampling the joints of people who had been eating it. For green tea nobody has, and until they do, a large and internally consistent chondrocyte literature stays exactly what it is: a good reason to run the study, not a reason to drink the tea.