There is a moment familiar to almost every family living with a dementia diagnosis. It happens in a kitchen, usually early, with the weekly pill organiser open and the small white tablet sitting in the Tuesday compartment. Someone asks the question out loud for the first time: is this actually doing anything any more? And behind it sits a second question, sharper in households where the letter from the memory clinic said not Alzheimer's disease alone but mixed dementia, Alzheimer's changes alongside damage from blood vessel disease. The drug trials that licensed these tablets largely did not enrol people like this. So what is the tablet doing, in someone the trials never studied?
On 28 November 2025, in Alzheimer's Research and Therapy, Cen Chen, Minjia Mo, Madeleine Akerman, Sara Garcia-Ptacek, Hong Xu and Maria Eriksdotter of the Karolinska Institutet published the largest attempt yet to answer that from real practice. Using SveDem, the Swedish registry for cognitive and dementia disorders, they took 32,282 people diagnosed with Alzheimer's disease or mixed dementia between 2007 and 2022, mean age 78.4 years, 61 per cent women, and followed them for up to 11 years. This is an observational study, not a randomised trial: nobody was assigned a drug for research purposes. What the team did was model treatment as it actually changes over time, checking within six months of each follow up whether a person was on a cholinesterase inhibitor alone, memantine alone, or memantine added on to a cholinesterase inhibitor, and matching that against their Mini Mental State Examination scores, the 30 point test used at Swedish follow up visits. Of the cohort, 78.4 per cent started on a cholinesterase inhibitor alone (donepezil, galantamine or rivastigmine) and 21.6 per cent on memantine alone.
The headline numbers are these. People on memantine alone lost 1.79 MMSE points a year (95 per cent confidence interval 1.73 to 1.85). Compared with them, people on a cholinesterase inhibitor alone declined 0.65 points less per year (0.59 to 0.72), and people on the combination declined 0.19 points less per year (0.10 to 0.28). Within the cholinesterase inhibitors, donepezil users declined 1.02 points a year (0.98 to 1.06); galantamine users lost an additional 0.05 points a year relative to donepezil, a difference whose confidence interval crosses zero and so cannot be distinguished from no difference, while rivastigmine users lost an additional 0.17 points a year (0.11 to 0.24). The authors also report that donepezil and galantamine users were more likely than rivastigmine users to reach one Defined Daily Dose, and that prescribing drifted over the years from single drugs towards combinations. Their own conclusion is measured: cholinesterase inhibitor users alone, particularly on donepezil and galantamine, showed slower decline, and the differences were modest.
Why this matters for the mixed group is a matter of who gets counted. In England, NICE guideline NG97, published on 20 June 2018, tells clinicians to consider these drugs in vascular dementia only where comorbid Alzheimer's disease, Parkinson's disease dementia or dementia with Lewy bodies is suspected, and not to offer them in frontotemporal dementia at all. The pure vascular case sits outside the licence; the mixed case sits in a grey zone that trials rarely illuminate. SveDem does not have that luxury, because it registers what Swedish clinics actually do, and mixed dementia is a large part of what they actually see. Putting Alzheimer's and mixed dementia into one 11 year model is therefore the point, not a compromise. It is also the study's boundary: the published figures describe the pooled group, not mixed dementia measured separately against pure Alzheimer's.
The caveat that a carer most needs to hold is about who ends up on what. Memantine is not a random alternative to a cholinesterase inhibitor. NICE positions memantine monotherapy for people with moderate Alzheimer's disease who cannot tolerate or must avoid the cholinesterase inhibitors, and for severe disease. In routine care, that means the memantine-alone group tends to be further along or less able to take the other drug, and a registry cannot fully scrub that difference out of the comparison. A gap of 0.65 MMSE points a year is not nothing across five years, but it is not a wall against the illness either, and it is an average across tens of thousands of people rather than a forecast for anyone. The same Swedish group's earlier work runs in the same direction with the same limits: in Neurology in 2021 they matched 11,652 cholinesterase inhibitor users against 5,826 untreated people with Alzheimer's dementia and found treated patients scored 0.13 MMSE points a year higher and had a 27 per cent lower risk of death (hazard ratio 0.73, 0.69 to 0.77), and in 2024 in Alzheimer's and Dementia they reported that in 1,095 people with dementia with Lewy bodies cholinesterase inhibitors, not memantine, tracked with slower decline.
And the third option, stopping? That has its own evidence, from elsewhere. In JAMA Network Open on 4 November 2024, Yu-Chien Lee and colleagues used Medicare claims from 2014 to 2019 in a propensity score matched cohort of older adults with dementia starting memantine, and found that stopping the cholinesterase inhibitor at that point was not associated with more days lost to long-term care institutionalisation over a year, and was associated with a lower risk of fall-related injury. That is a different question in a different health system, again observational. Taken together, what the last few years of registry work offers a family is not an instruction but a vocabulary: a sense of the size of the effects being argued over, the knowledge that mixed dementia is no longer simply missing from the evidence, and better questions for the next review appointment.