Adding Memantine to Donepezil, or Switching: A Moderate-Stage Crossroads
A patient on long-term donepezil now shows new nighttime agitation as his disease advances into moderate-severe stage. The disagreement isn't whether memantine has a role — it's whether it belongs added to what he's already taking, or in place of it.
Walter K., 81, spent forty years running a machine shop before he retired, the kind of work where a quarter-inch error mattered, and his wife says that precision is exactly what she misses watching disappear. He has stable hypertension and mild chronic kidney disease, otherwise healthy for his age, and was an avid woodworker until about a year ago, when he stopped recognizing which tools went where. He was diagnosed with Alzheimer's disease three years ago and has been on donepezil 10 mg for the past fourteen months, with an initial period where the family felt he was "holding steady." That impression has changed over the last six weeks: he has started getting up at 2 or 3 a.m., disoriented about where he is, occasionally trying to leave the house convinced he needs to "open the shop." His wife has started sleeping in shifts, and reports she is running out of ways to reassure him at 3 in the morning that the shop closed for good eleven years ago.
His MMSE has fallen to 12 from 17 a year ago, consistent with progression into moderate-to-severe disease — the stage at which memantine, an NMDA-receptor antagonist, has its own separate evidence base, distinct from donepezil's cholinergic mechanism. Memantine works by partially blocking excess glutamate signaling through NMDA receptors, a pattern of excitotoxic overactivity that becomes more prominent as disease advances; the question in front of the team isn't whether memantine has earned a place in his regimen at this stage — both clinicians in the room agree it has — but whether it should be added to the donepezil he's already on, or whether the donepezil, whose apparent benefit his wife describes as having quietly faded months before the new agitation appeared, should be stopped and replaced.
At the fourteen-month follow-up
I'd add memantine to his current donepezil rather than switch. The relevant trial here is a 404-patient, placebo-controlled study — Tariot and colleagues, published in JAMA in 2004 — that specifically randomized patients already stable on donepezil to add either memantine or placebo. The combination group showed significantly better outcomes on the Severe Impairment Battery for cognition, on activities-of-daily-living scores, and on the Neuropsychiatric Inventory for behavioral symptoms — the exact domain we're worried about tonight.
I'd point to a different trial with a harder outcome. DOMINO-AD — Howard and colleagues, New England Journal of Medicine, 2012 — used a two-by-two design in 295 patients with moderate-to-severe disease already on donepezil, randomizing to continue or discontinue donepezil, and separately to start or not start memantine. The nursing-home-placement result came later, in the trial's secondary and post-hoc analysis — Howard and colleagues again, Lancet Neurology, 2015: continuing donepezil roughly halved the risk of nursing-home placement over the first year. Adding memantine, in that same analysis, made no measurable difference to that outcome, with or without donepezil on board.
I'm not disputing the Tariot data on cognition and behavior scores — I'm saying that on the outcome that actually determines whether Walter stays in his own house, the trial that tested it directly found memantine added nothing. If we're choosing what matters most tonight, keep the donepezil.
I don't think either trial answers tonight's actual question, which isn't "does combination beat monotherapy on average" — it's why is he getting up at 2 a.m. newly disoriented, six weeks ago and not before. That kind of nighttime agitation in advancing disease tracks with the glutamatergic excitotoxicity memantine's NMDA-receptor blockade is specifically built to address, distinct from donepezil's cholinergic target.
You're both citing population averages from trials that weren't designed around a new, discrete symptom like this one. I'd keep the donepezil — there's no switch-benefit evidence to justify losing whatever cholinergic support he still has — and add memantine now at the standard target dose, specifically because his own presentation, not the trial averages, points there. And to be precise about his kidneys, since it comes up every time an eGFR in the fifties appears on a chart: memantine only needs its target dose halved in severe renal impairment, a creatinine clearance of 5 to 29. At 52 he is nowhere near that, so he gets the full 10 mg twice daily — reflexively cutting it here would under-treat him for no reason.
Agreed: continue donepezil 10 mg unchanged, add memantine at 5 mg daily with titration to the full 10 mg twice-daily target over four weeks — no renal dose reduction, his eGFR of 52 being well above the severe-impairment threshold that would call for one — and reassess the nighttime symptoms specifically, not just global cognition, at six weeks.
Not agreed: how much weight to give the DOMINO-AD null finding on memantine's added value going forward, once Walter's own nighttime symptoms are no longer the acute question. The geriatrician wants that trial's caution to inform future deprescribing conversations as his disease advances further; the neurologist views today's decision as symptom-specific and not a comment on memantine's aggregate value one way or the other. Both left the point on record rather than resolving it.