Milk & dairy

Milk has a long association with bones and almost none with cartilage. What exists is one striking dataset, one careful contradiction of it, and a set of nulls — and the interesting part is how a result can be statistically clean and still not be worth much.

The Osteoarthritis Initiative result

Two thousand one hundred and forty-eight people with radiographic knee OA, 3,064 knees, followed for four years, with joint space width measured quantitatively from plain films rather than graded by eye. Among women, mean loss of medial joint space ran in a straight line against baseline milk intake: 0.38 mm at no milk, 0.29 mm at three glasses a week or fewer, 0.29 mm at four to six, and 0.26 mm at seven or more (p = 0.014 for trend), adjusted for baseline severity, body mass index and other dietary factors.

Among men, there was no association at any level of intake.

Two things are true about that at once. It is a proper prospective analysis with a quantitative structural endpoint, which is rare in nutrition research and is why it gets cited. And the entire spread, from drinking no milk to drinking a glass a day, is 0.12 mm of joint space over four years — a difference sitting close to the measurement's own resolution, present in one sex, absent in the other, with no mechanism offered for the split, and never replicated in the twelve years since despite the dataset staying open. Findings that appear in women and vanish in men are as often a feature of subgroup analysis as of biology.

The soy milk study, and the split it reports

In 5,764 adults in Hunan with radiographs scored on the OARSI atlas, soy milk intake tracked with fewer osteophytes in a dose-response — at least daily, OR 0.49 (95% CI 0.26–0.92); less than daily, OR 0.79 (0.67–0.94); p for trend 0.001 — surviving adjustment for energy intake, activity, smoking, alcohol, education, diabetes, hypertension and other dairy consumption.

And joint space narrowing showed no association at all, in either model.

That split is the most informative thing in the topic. Osteophytes are new bone at the joint margin; joint space narrowing is the cartilage going. A food associated with the first and not the second is telling you something about bone remodelling, not about cartilage — and a study that reports both separately, instead of collapsing them into a single OA grade, is the reason we can say so.

Diet quality against MRI

Three hundred and ninety-two people with symptomatic knee OA were followed for two years with MRI: cartilage volume, cartilage defects, bone marrow lesions, effusion-synovitis. Diet quality — scored across vegetable, fruit, grain, dairy, fat and alcohol sub-scores — was associated with none of them, nor with symptoms. It was associated with more lower-limb muscle strength, fewer depressive symptoms and better quality of life, and among the food groups only the vegetable sub-score reproduced that pattern. The dairy sub-score carried nothing.

In the OAI imaging sub-cohort, no dietary variable predicted twelve-month change in effusion-synovitis. And a Mendelian randomization did find genetically predicted cheese intake associated with lower knee OA risk — with the authors' own multivariable analysis routing the dietary effects through obesity, BMI, diabetes and hypertension rather than through anything in the cheese.

The oldest signal, from 655 people interviewed in Antalya in 2004, found symptomatic knee OA less common among daily milk drinkers and among tea drinkers, with nothing else in the diet showing an association. Cross-sectional, unadjusted for weight, and prevalent disease — people whose knees already hurt have had time to change what they eat.

The probiotic trial

Eighty people took a Streptococcus thermophilus strain — a yogurt bacterium, isolated from human breast milk, that produces hyaluronate in the gut — or placebo for twelve weeks. Serum CTX-II, a marker of type II collagen breakdown, and CRP both fell in the probiotic group. The authors report that the improvement was indistinct against placebo and attribute that to uneven disease severity between groups, the short period and the small sample. Six of the eight authors work for the company that owns the strain.

It is worth knowing about mostly because of what it is not: a trial of dairy.

The mechanism this topic does not have

No dairy-specific cartilage mechanism appears anywhere in this literature. The plausible routes all belong to other entries — calcium and vitamin D acting on subchondral bone, protein supporting the muscle that unloads the joint, menaquinone in fermented dairy acting through matrix Gla protein, and fermented dairy acting on the gut microbiome. That absence is not an argument against milk. It is the reason a 0.12 mm association has nowhere to go.

What would change this entry

Replication, in the other sex or in another cohort, with a cartilage-specific measure. The OAI has dGEMRIC and T2 mapping in sub-cohorts; running the milk analysis against cartilage composition rather than radiographic joint space would say in one study whether there is anything here worth pursuing.