Deprescribing Dementia Medications: Two Patients, Two Different Answers
Two families ask the same question about the same drug class at an advanced stage of the same disease. The pharmacology doesn't change between them — what changes is which outcomes the evidence was ever built to measure, and whether those are still the outcomes that matter.
Dorothy M., 86, was a church organist for over fifty years and, until eighteen months ago, could still pick out hymns from memory even after conversation had become difficult. She has advanced Alzheimer's disease, MMSE un-scoreable for over a year, but she remains ambulatory with assistance, recognizes her daughter's voice, and still smiles reliably at music. She lives in a memory-care facility and her daughter visits three times a week.
She has been on donepezil and memantine for six years. Her daughter, prompted by a friend whose mother's medications were "cleaned up" at a similar stage, asks today whether Dorothy still needs these drugs, or whether continuing them is really just inertia. There is no acute problem driving the question — no new swallowing difficulty, no medication burden crisis, no shift in goals of care. Dorothy's baseline, for her stage, has been stable for months, and the family's goals remain what they have been throughout: maximize her remaining function and quality of interaction for as long as that is realistic.
I wouldn't stop either drug today. The most relevant trial here is DOMINO-AD — Howard and colleagues, New England Journal of Medicine, 2012 — 295 patients with moderate-to-severe Alzheimer's disease already on donepezil, randomized to continue or discontinue. The nursing-home finding comes from the trial's secondary and post-hoc analysis, published separately in Lancet Neurology in 2015: discontinuation significantly increased the risk of nursing-home placement within the first year, hazard ratio 2.09, and that difference didn't persist over the following three years.
Dorothy's situation — stable function, no new problem, family simply asking on principle — is almost exactly the population that trial enrolled. I don't see a reason today to deviate from what it found.
I'm not proposing we stop today either, but I want to name the limit of that trial honestly: it tested one year of follow-up in patients who'd been on donepezil a median of well under six years. We don't actually know what the risk-benefit balance looks like this far out, and "we've never re-checked in six years" is a real gap on its own, independent of what DOMINO-AD found.
I'd agree to continue today, but I'd want that gap named explicitly rather than treated as resolved by a trial that wasn't really testing Dorothy's specific six-year scenario.
Agreed: continue both medications unchanged. The daughter's question is documented as answered for today, not dismissed — the team commits to revisiting the decision explicitly if Dorothy's function, swallowing, or goals of care change.
Not agreed: whether six years on these drugs, on its own, should eventually trigger a scheduled reassessment independent of any new clinical event. The pharmacologist wants a standing checkpoint; the geriatrician sees no evidence-based interval to set one by and prefers to let clinical change, not the calendar, drive the next review.
Harold P., 89, was Dorothy's exact contemporary in disease stage eight months ago; the two families in fact met once in the same waiting room. Since then Harold's course has diverged sharply. A hospital admission for aspiration pneumonia three months ago left his family and his physicians in agreement, after a long conversation, that his goals of care should shift to comfort-focused: no further hospitalization, no feeding tube, care aimed at minimizing distress rather than extending function. He is now enrolled in hospice.
He remains on donepezil and memantine, continued out of habit through the hospitalization and hospice enrollment rather than by any deliberate decision. He now has documented swallowing difficulty significant enough that his hospice nurse reports his oral medications, memantine in particular, are frequently followed by coughing and visible distress at administration. His daughter, present for today's hospice recertification visit, says plainly that watching him choke on pills twice a day, for medications she no longer understands the point of, is one of the hardest parts of her day.
For consistency with what I argued in Dorothy's case, I want to name that the same DOMINO-AD signal exists here too — there's some real, if small and unquantified, risk that stopping abruptly could accelerate decline in a way none of us would want to see in his remaining time.
I'm not going to argue hard for continuing, though. I think this case is different enough that I'd defer to palliative care's framing below.
The DOMINO-AD outcome set — nursing-home placement, functional trajectory over a year — simply isn't the outcome set that governs Harold's care anymore. His family made an explicit, documented decision six weeks ago to shift toward comfort-focused care specifically because further aggressive management, including another hospitalization, was no longer what they wanted for him.
The concern about destabilizing him is worth naming, but the medication itself is causing a form of distress — coughing, visible discomfort twice a day — that comfort-focused care exists specifically to prevent. Continuing it isn't neutral; it's working against the goals his family already set.
Agreed: both donepezil and memantine are tapered and discontinued, with the family informed this reflects the goals-of-care shift already made, not a withdrawal of care. Hospice nursing will monitor for any new agitation or decline in the two weeks following discontinuation.
Not agreed, and left explicitly on record: whether the same discontinuation-risk data should have been mentioned to the family at all today, given how little it seemed to actually apply. The geriatrician felt naming it was honest practice; the palliative care specialist worried it could needlessly reopen a decision the family had already made peace with.