What the Pills Actually Do

Nobody sits you down and tells you how small the effect is. Knowing the real size of it changes what you spend your hope on.

Do donepezil and memantine actually work?

Yes, measurably — and by less than almost everyone imagines. The average benefit sits just below the line usually drawn for a clinically important change.

The cognitive scale used in most dementia drug trials is the ADAS-Cog. On that scale, a 4-point change over six months is the figure conventionally treated as clinically important.

Pooled across five randomized trials, donepezil at 10mg daily produced a weighted mean difference of 2.80 points versus placebo.

Below the line. Not zero — but below the line.

An evidence review conducted for a clinical practice guideline put it in one phrase: cholinesterase inhibitors and memantine produce improvements in cognition and global assessment that are statistically significant but clinically marginal.

Memantine works differently — it acts on glutamate rather than acetylcholine — and is used in moderate to severe disease, often alongside donepezil. Adding memantine to donepezil has generally not shown significant benefit over donepezil alone.

Then why is he on it?

Because a small average effect is not the same as no effect, and because for some individuals the effect is larger than the average.

Averages hide the spread. Some people respond noticeably; some not at all. The trial result is the middle of that range, not a prediction about your person.

There are reasonable arguments for staying on these drugs:

• Some families do see a difference — more alertness, more engagement, a slower slide — and that is worth having.
• In moderate to severe disease, continued donepezil has been associated with functional benefits over a year, not only cognitive ones.
• Stopping can produce a drop that does not fully recover when the drug is restarted. That risk shapes a lot of prescribing.

And there are reasonable arguments against continuing indefinitely, which is the subject of its own conversation about deprescribing — because these medications carry real side effects, and in late disease the reason for taking them can quietly expire while the prescription does not.

What is not reasonable is being told "this will help" and inferring that the disease is being treated. It is being nudged.

What about Leqembi and Kisunla?

They do something the older drugs do not — they clear amyloid and slow the rate of decline. Whether the amount of slowing is meaningful to a family is genuinely contested.

The anti-amyloid antibodies — lecanemab (Leqembi) and donanemab (Kisunla) — represent the first treatments that alter the disease process rather than the symptoms. Both produced statistically robust effects on cognitive and functional decline in early Alzheimer's.

The debate is about size. Donanemab slowed decline on the CDR-SB scale by roughly 22 percent in the overall trial population, and by about 35 percent in participants with intermediate tau levels. In a European regulatory assessment published this year, the reviewers accepted the effects as methodologically robust while noting that expert opinion on clinical meaningfulness "considerably differed" — and that published estimates of the minimum clinically important difference are higher than what these drugs achieved.

They also carry a specific risk: ARIA, amyloid-related imaging abnormalities — brain swelling and small bleeds found on MRI. Most are asymptomatic; some are serious. Rates differ between the two drugs, and they differ sharply by genotype. With standard donanemab dosing, ARIA-E occurred in about 24 percent overall and 57 percent of people carrying two copies of APOE4; a modified titration schedule brought those figures down to roughly 14 percent and 19 percent.

Eligibility is narrow, the infusions and MRI monitoring are demanding, and this is early-disease treatment only. We cover who qualifies and what the process involves separately.

Is there anything that helps the behavior?

Yes, and the best-evidenced options are not medications.

This is the request most families actually arrive with. Not "make her remember" but "make the nights survivable."

A Bayesian network meta-analysis of 65 randomized trials compared eleven non-drug approaches to agitation. The ones associated with the largest reductions were personally tailored interventions and massage and touch therapy, with animal-assisted intervention also ranking highly. Non-drug approaches are recommended first-line in essentially every guideline, partly because they work and partly because the drug alternatives are not benign.

When medication does come into it, the numbers deserve saying plainly. Across 17 placebo-controlled trials in 5,377 older adults with dementia-related behavioral disturbance, mortality was about 4.5 percent on antipsychotics versus 2.6 percent on placebo — a relative risk of roughly 1.6 to 1.7. That finding is why the FDA applied a boxed warning to atypical antipsychotics in 2005, and extended it to first-generation antipsychotics, which carry higher risk still, in 2008.

That is not an argument that antipsychotics are never appropriate. Severe distress, psychosis and danger are real, and sometimes the alternative is worse. It is an argument that this is a serious decision with a mortality figure attached, and it should be made deliberately, at a defined dose, with a defined review date — not drifted into and left running for years.

What has better evidence than most of the drugs?

The unglamorous things. Pain, hearing, sleep and daylight change behavior more reliably than most prescriptions do.

Four that are consistently under-treated:

• Pain. Untreated pain is one of the most common drivers of agitation in people who can no longer report it in words. A person who becomes "aggressive during personal care" is very often a person whose shoulder hurts. Scheduled analgesia — tried properly, not as-needed — is one of the highest-yield experiments available.
• Hearing. Uncorrected hearing loss produces confusion, withdrawal, suspicion and apparent paranoia. Hearing aids that sit in a drawer help nobody; a working amplifier that gets used is worth more than a dose adjustment.
• Sleep and daylight. A predictable day with real light exposure in the morning is not a soft intervention. It is one of the few things that reliably shifts a disordered night.
• The rest of the medicine cabinet. Anticholinergic drugs — some bladder medications, some sleep aids, some antihistamines — actively worsen cognition, and many people with dementia are on more than one without anyone having added them up. Reviewing that list often produces a bigger change than adding anything new.

There is something bleak in the fact that the highest-value interventions are also the ones nobody profits from, and so nobody markets them. But that is how it is.

How do I tell whether a medication is doing anything?

Pick two or three concrete things you can observe, write down where they stand today, and look again in eight to twelve weeks.

Nobody can answer "is it working" from memory, because the comparison you need is against a version of the person you can no longer accurately recall.

Before a new medication starts, or a dose changes, write down:

• Two specific daily things. Not "memory" — "can find the bathroom at night," "starts a conversation at dinner," "gets dressed with one prompt instead of five."
• A frequency. How many nights a week is he up. How many times a week does personal care end in a fight.
• The date, and what was changed.

Then leave it alone for eight to twelve weeks unless something goes wrong, and read it back. A written record is also what protects you from the two failure modes: stopping something useful because a bad couple of weeks coincided with it, and continuing something useless for years because nobody set a date to look.

What should I ask at the next appointment?

Ask what the medication is expected to do, how you would know if it were working, and when it will be reviewed.

Four questions that tend to produce more honest answers than "is there anything else we can try":

• "What specifically should I expect to see, and roughly how big is that effect likely to be?"
• "How will we decide whether it is working, and when will we look?"
• "What are the side effects worth calling you about, and which ones tend to be mistaken for the dementia getting worse?"
• "Is there anything on this list that could come off?"

That last one is often the most productive question in the whole visit.

If you take one thing from this: the medications are worth trying and worth reviewing, and they are not the main lever. The main lever is the pain nobody has looked for, the hearing aid nobody replaced the battery in, and the drug already on the list that is making everything worse.