What the study found
Researchers at the University of Florida used artificial intelligence to comb deidentified health records from 2012 to 2024, looking for patients diagnosed with Alzheimer's disease or related dementias, or with mild cognitive impairment.
About 8% of both groups reported taking glucosamine, a popular over-the-counter supplement used for joint pain: 1,896 patients with dementia and 2,750 with mild cognitive impairment.
After controlling for age, sex and demographics, the analysis showed glucosamine use was associated with a 25% higher likelihood that mild cognitive impairment progressed to dementia. A separate look at records from nearly 42,000 people with mild cognitive impairment found a similar pattern.
Association is not proof
This is observational data drawn from medical records, not a randomised controlled trial. It shows a link between glucosamine use and faster progression. It does not show that the supplement causes the change.
People who take glucosamine may differ in ways the records do not capture. Joint pain itself, and the arthritis that drives it, could be markers tied to other health factors. Supplement use was self-reported, and people who take one supplement may take others.
Researchers also ran laboratory experiments to look for a biological mechanism, which is what turns a statistical signal into a testable hypothesis. That work is early.
What patients should take from it
Mild cognitive impairment describes measurable problems with memory or thinking that go beyond normal ageing but do not substantially interfere with daily life. A meaningful share of people with the condition never progress to dementia.
The practical advice from clinicians is not to change a medication or supplement routine on the basis of one observational study. Anyone taking glucosamine for joint pain who has memory concerns should raise both with a doctor, who can weigh the arthritis benefit against the new signal.
The findings add to a growing list of common supplements that look harmless in the bottle but interact with conditions in ways that only large record analyses can see.