Oily fish

The advice to eat more oily fish for your joints is old enough that most people have stopped asking where it came from. It came from capsules — from trials of 2 to 4.5 grams a day of concentrated marine oil, which is not what a fillet delivers. This entry is about the fish. The dosed-oil trials have their own entry, and the two bodies of evidence do not agree.

What the fish-eating cohorts found

Thirty-five thousand people in the Melbourne Collaborative Cohort had their diets recorded between 1990 and 1994, and their records were later linked to the Australian National Joint Replacement Registry — a hard endpoint, an operation rather than a questionnaire. Fish intake was associated with neither hip nor knee replacement. The same analysis found a dose-response inverse association between fresh red meat and hip replacement (HR 0.94 per extra serving per week, 95% CI 0.89–0.98), which is a useful calibration for how much weight to put on any single food-and-joint association from this kind of data.

In 659 knees from the Osteoarthritis Initiative imaged for effusion-synovitis, higher intake of meat, fish, poultry, eggs and legumes was associated with larger baseline synovitis volume (β = 0.452, 95% CI 0.075–0.829, p = 0.019) — an association that did not survive false-discovery correction. Over twelve months, no dietary variable at all predicted change in synovitis, across every sensitivity analysis the authors ran.

A Mendelian randomization of eighteen dietary factors did find genetically predicted oily fish intake associated with lower knee OA risk. Its own authors report that the effect ran through obesity, body mass index, diabetes, hypertension and prolonged standing — which is a statement that the route is metabolic rather than cartilaginous.

The trial that changed the picture

For years the summary of marine oil in osteoarthritis was straightforward: good mechanism, null trials. That is no longer quite true. In a six-month, multicentre, double-blind trial, 235 adults with mild-to-moderate knee OA took 4 g/day of krill oil — 0.60 g EPA, 0.28 g DHA, 0.45 g astaxanthin — or a vegetable-oil placebo. Their Omega-3 Index rose from 6.0% to 8.9% while placebo stayed flat. Knee pain improved in both groups and more on krill oil: an adjusted between-group difference of −5.18 on a 0–100 scale (95% CI −10.0 to −0.32, p = 0.04), with stiffness and physical function moving the same way.

Five points on a hundred-point scale, with an interval reaching −0.32, is a small effect measured precisely. And the inflammatory markers — the proposed mechanism — did not move at all.

Who was short to begin with

The detail in that trial worth carrying away is an entry criterion. Participants had to be consuming less than 0.5 grams a day of long-chain omega-3s to qualify. It was, by design, a repletion study: it enrolled people who were short and made them not short.

That is the most plausible reconciliation between this trial and the ones that found nothing, and it changes what the question is. If the effect is repletion rather than pharmacology, then it applies to people whose Omega-3 Index is low and not to everyone — and someone eating oily fish twice a week has already had the intervention. Nobody has yet run a trial stratified on baseline Omega-3 Index, which is odd, because every trial now measures it.

Two smaller records fill in the edges. A Japanese trial reported knee-pain improvement from 149 mg of EPA — a twentieth of the OA trial doses — but it was given with l-serine, so nothing in the design attributes the result to EPA, and every author but the clinician was employed by the company selling it. And in 249 women with aromatase-inhibitor joint pain, 3.3 g/day of omega-3 halved pain scores — as did placebo, with no difference between them (p = 0.58), while the fall in triglycerides confirmed the capsules were being taken.

The mouse that makes its own

Because of a mouse that makes its own. The fat-1 transgenic mouse converts n-6 fatty acids to n-3 endogenously, so its tissues carry a different fatty-acid composition without anything being fed to it. After meniscal resection, those mice had significantly less cartilage destruction and osteophyte formation (p < 0.01), lower MMP-13 and ADAMTS-5 protein in articular cartilage (p < 0.01), and less chondrocyte and matrix loss, through mTORC1 downregulation and increased chondrocyte autophagy.

That is a study about tissue omega-3 status rather than about a supplement, and it is the cleanest version of the hypothesis anyone has tested. It is also a mouse whose lifelong fatty-acid composition no diet reproduces.

Practical notes

The doses that have shown anything in a knee are around 0.6 g/day EPA plus 0.3 g/day DHA. Portion-level arithmetic — how many meals of which species that represents — is not something this entry will assert, because no indexed source tying portion composition to a joint endpoint turned up in the search behind it.

Methylmercury accumulates in large predatory fish, which is why public advice distinguishes species and limits portions for some groups. That precaution stands on its own evidence; nothing links mercury exposure to a joint outcome, and this entry will not imply that it does.

What would change this entry

A trial of oily fish as food, with a joint endpoint — which has never been run in any population. Failing that, a marine-oil trial stratified by baseline Omega-3 Index would answer the more interesting question the krill trial raised by accident: not whether omega-3s work, but whether they work for people who already have enough.