Donepezil for Mild Alzheimer's: Weighing a Modest Effect Size
A newly diagnosed patient's family wants to start a cholinesterase inhibitor immediately. The disagreement isn't about the diagnosis — it's about how much weight a real but genuinely small measured benefit should carry against the momentum of a family already committed to "doing something."
Eleanor R., a 78-year-old retired reference librarian, sits with her husband and her daughter Grace while the diagnosis is confirmed: probable Alzheimer's disease, mild stage. She has been generally healthy — well-controlled hypertension on lisinopril for over a decade, no falls, no cardiac history — and still tends an ambitious vegetable garden every summer, though her husband mentions she forgot to harvest most of it this past August. The family describes eighteen months of a slow, specific decline: missed bill payments that were never missed before, the same story told three times at dinner, and one frightening afternoon last spring when she got lost driving home from a grocery store she has shopped at for thirty years.
Grace has already read about donepezil online before this appointment and arrives with a folder of printouts. She wants it started today. Her urgency is not unreasonable — watching a parent's memory erode invites exactly this response — but the actual evidence for a cholinesterase inhibitor in mild Alzheimer's disease is smaller and more contested than the framing she has absorbed. Cholinesterase inhibitors raise synaptic acetylcholine by blocking its breakdown, partially compensating for the cholinergic neuron loss that tracks disease progression; the pharmacology is sound, but soundness of mechanism and size of measured clinical benefit are two different questions, and the team now has to decide how to answer the second one honestly, in the room, with a family already emotionally committed to a specific next step, a family that has clearly spent the drive here rehearsing what they wanted this appointment to accomplish.
In clinic, after the diagnosis is given
I'd start donepezil today, at 5 mg with a titration to 10 mg at four weeks. The Cochrane review of cholinesterase inhibitor trials in Alzheimer's disease — pooling roughly two dozen randomized trials — found a consistent benefit of about 2.7 points on the ADAS-Cog scale relative to placebo. That's small on a 70-point instrument, I know that going in, but it has replicated across a genuinely large and heterogeneous trial base, and the American Academy of Neurology's own practice guideline recommends offering a cholinesterase inhibitor at diagnosis on that basis.
I'd rather give her family a real, if modest, intervention now than tell them to wait for something better that may not come.
The Cochrane pooling is real, and I'm not disputing the 2.7-point figure. But the trial that actually tested whether that translates into anything a family would notice is AD2000 — a large, pragmatic, UK-based randomized trial of donepezil in mild-to-moderate Alzheimer's disease, run specifically because the earlier manufacturer-sponsored trials hadn't looked at institutionalization or disability progression as primary outcomes. AD2000 found no significant difference in time to institutionalization or in the rate of disability progression over three years, despite showing the expected difference on cognitive and behavioral rating scales.
A 2.7-point difference on a 70-point scale is below what most of us can detect at the bedside without the test in front of us. I'm not saying the drug does nothing — I'm saying the outcome AD2000 was designed to test, the one closest to what Grace actually wants to know, didn't move.
You're both right about the numbers, and I don't think this family needs to hear us resolve which trial matters more — they need a plan. Here's what I'd actually propose: start donepezil, titrate as above, but set an explicit four-month reassessment using a caregiver-rated global impression, not just an office MMSE. If Grace and her father can't identify anything meaningfully different by then, we stop it — not "continue because we already started," which is what happens to most patients on this drug in practice.
The AD2000 finding is exactly why I want a real stopping rule, not a reason to skip the trial. If institutionalization risk truly isn't moving, we shouldn't be keeping her on a drug indefinitely on faith that it's doing something we can't measure.
Agreed: donepezil started today at 5 mg, titrated to 10 mg at four weeks, alongside a referral for cognitive stimulation therapy. A caregiver-rated global-impression reassessment is scheduled explicitly for four months from today, documented in the chart as a decision point, not a routine follow-up.
Continue at the current dose and reassess again at the next annual visit, same standard.
Taper and discontinue rather than continue indefinitely on the strength of the diagnosis alone.
Not agreed: whether AD2000's institutionalization finding should have been raised with the family today at all, or saved for the four-month conversation if the drug appears not to be helping. The psychiatrist worried that raising it now would undercut a fragile family's willingness to try; the pharmacologist thought withholding it was a form of the same momentum problem the reassessment plan was built to counter. Both views were left on record, unresolved.