Glucosamine dulled memory in Alzheimer's mice; the evidence in people is thinner

In a Nature Metabolism paper, two weeks of the joint supplement worsened a memory test in mice bred to model Alzheimer's (7 treated, 6 controls). In UF Health records, people with mild memory problems whose notes mentioned glucosamine were somewhat more likely to be diagnosed with dementia later, but the comparison allowed only for demographics, not for why people take it, and a first independent check didn't single out glucosamine.

8 October 2026 · How we rate evidence

Mice don't have arthritis clinics, so the University of Florida team had to put the glucosamine in by tube: 457 milligrams per kilogram of body weight a day for two weeks. Scaled to a person by the usual body-surface formula, that is about 2,600 mg a day, well above the 1,500 mg a day given in GAIT, a US trial of 1,583 people with knee arthritis. The mice getting it were bred to develop Alzheimer's-like brain changes. After two weeks, they were worse at recognising a mouse they had already met.

The claim going around. "Glucosamine, a popular joint supplement, linked to faster Alzheimer's progression," ScienceDaily headlined on 27 September, reprinting a UF Health release: "a 25% higher likelihood of mild cognitive impairment progressing to dementia." In June, Sun had written in The Conversation that people with mild impairment taking glucosamine "were 25% more likely to progress to full Alzheimer's."

What they did. Glucosamine is a side plot. The paper, led by Ramon Sun, Matthew Gentry and Craig Vander Kooi, is about sugar chains called N-glycans, which cells attach to many proteins to help them fold and reach the right place. Using a mass-spectrometry method that maps molecules across a slice of brain, the team found more of these chains in Alzheimer's brains (three patients against three controls) and in two mouse models. Tracing labelled glucose suggested the mice were making glycans faster, not clearing them more slowly.

Then, in the paper's cleanest test, they pushed glycan production both ways. Turning it down, by silencing the enzyme PGM3 or with an experimental drug, NGI-1, improved the mice's social memory. Turning it up with glucosamine, which the body feeds into the same assembly line, made it worse. Software scored behaviour; no mice were excluded.

Finally they looked for glucosamine in people, searching doctors' notes at UF Health for the word. About 7 to 8% of patients with mild cognitive impairment or dementia had it mentioned. Each user was paired with a non-user of similar age and, in some analyses, sex and ethnicity.

What they found. In people with mild cognitive impairment, glucosamine was mentioned alongside more later diagnoses of dementia. Reading the paper's plotted data, the two groups were level for the first year. By five years, about 28 in 100 users had been diagnosed, against 24 in 100 non-users. By ten years it was 62 against 50, among the roughly 600 people per group still being followed. The "25%" is that ten-year gap in relative terms (62 divided by 50), not 25 percentage points. The paper gives no hazard ratio, and doesn't name the test behind its P value.

Among people already diagnosed with dementia, 49 in 100 glucosamine users died within ten years, against 37 in 100 matched non-users; at five years it was about 36 against 31. Among people with mild impairment there was no difference in survival, and users did slightly better for the first three years. The authors read that split as glucosamine harming only a brain that is already sick. A plain alternative is that the comparison is uneven, in ways that play out differently in each group.

What it doesn't show. The health-record analysis can't tell glucosamine apart from the reasons people take it. Glucosamine is taken for joints, but arthritis, painkiller use, weight, diabetes, heart disease and how often people see a doctor were not accounted for. Only age and, at most, sex and ethnicity were. The outcome, a dementia diagnosis in the same records, is more likely to be written down for people who visit more often, and people with more visits also have more notes where "glucosamine" can turn up. The paper says use was documented for at least a year after diagnosis, yet users with dementia died in that first year as often as non-users, so how that rule was applied is unclear.

A few details in the paper don't match each other. The text says about 8% used glucosamine; its figure shows 7.6% for mild impairment, its counts 6.6%. The methods say only female mice and no randomisation; the reporting summary says both sexes, randomly assigned.

The mouse work is small: 7 treated and 6 control mice for glucosamine, a single memory test, and a difference that showed only on the fourth trial. Neither lowering nor raising glycans changed amyloid plaques or inflamed support cells (3 to 5 mice per group); lowering didn't change tau in the tau model. Our rating of the lowering result, Speculative or Early, turns on whether groups that small could have seen such a change.

What it means beyond the study. The human question is whether glucosamine, at ordinary doses, speeds decline in someone whose memory is already slipping. This paper can't answer it, and the first outside check doesn't help the case. In July, a team at the University of Tokyo posted an analysis of the US National Alzheimer's Coordinating Center cohort. People with mild impairment and a glucosamine record progressed to Alzheimer's slightly more often (hazard ratio 1.16, range 0.98 to 1.37, not clear of chance); for any dementia, the outcome UF counted, it was 1.07 (0.92 to 1.24). Similar estimates turned up for multivitamin and calcium/vitamin D records. Among people with dementia, glucosamine went with no higher death rate (0.93, range 0.83 to 1.05). That analysis is not yet peer-reviewed. Earlier UK Biobank studies (2022, 2024) of people without dementia found glucosamine users had the same or slightly lower dementia rates; Sun argues glucosamine may act differently once the brain is already declining.

The ScienceDaily text is careful about the health records. It says the records "do not show that glucosamine itself causes dementia to progress," and quotes Gentry calling them "an association and not proof of causality." The headline's "Alzheimer's progression" is looser: the outcome counted any dementia, including vascular, Lewy body and frontotemporal.

Why it's still interesting. Alzheimer's research has spent decades on plaques and tangles. This paper points at something different: the brain's protein-decorating machinery running too hot, and a memory deficit in mice that moved when the machinery was turned down, even though plaques didn't budge. If that holds in more mice, over longer treatment, it is a new kind of drug target. Glucosamine was the team's tool for turning the dial up, and it did what the idea predicted in mice.

Open questions.

Is the signal glucosamine, or the kind of person whose notes mention it? Re-run the health-record analysis with arthritis, painkillers and visit counts accounted for, and set glucosamine beside other supplements recorded the same way. If glucosamine stands out, it earns a trial. If every supplement looks alike, the records are describing patients, not a pill.

Does a normal human dose change the brain? The mice got the equivalent of about 2,600 mg a day. Measure glycan markers in spinal fluid in people before and after taking 1,500 mg a day. If nothing moves, the mouse mechanism has no route to matter for ordinary users.

Are extra glycans a cause or a response? The authors themselves suggest the build-up might be the brain compensating. Test whether it appears before memory loss in mice, and whether lowering it early helps while lowering it late harms. The answer decides whether blocking it is a treatment or a mistake.

A note on timing. The paper came out on 9 June 2026, and Sun's Conversation piece appeared the same day. The ScienceDaily repost arrived on 27 September, three and a half months later. By then the independent preprint had been online for two months.