Donepezil and Memantine: Independent Evidence, Risks & Funding

Table of contents

• Evidence summary
• What these medicines are
• How they work
• What they are prescribed for
• What works and what does not
• Benefits by claim
• Risks and all side effects
• All interactions
• Who should avoid them
• Dosage and how to take them
• Follow the money: who makes them and who funded the evidence
• Related research
• Frequently asked questions
• Sources and funding notes

Evidence summary

Independent evidence and credibility scorecard

Independence tiers: 1 = regulator, government body, or publicly or charity-funded work with no declared commercial funding; 2 = academic or Cochrane team with no declared commercial funding for the work, but analysing mostly sponsored trials or with minor industry ties; 3 = authors with significant declared industry ties; 4 = manufacturer-sponsored. Credibility runs from A (highest) to D.

The independent sources cluster in the UK (seven of twelve rows). That means the verdict reflects the NHS-and-NICE lens, in which cost-effectiveness is central. The sources with opposing interests, the public AD2000 trial and the maker-funded trials, agree on a small cognitive gain but differ on whether it matters in daily life.

What these medicines are

• Donepezil (brand Aricept): a cholinesterase inhibitor in 5, 10 and 23 mg tablets and dissolving tablets (US label).
• Rivastigmine (brand Exelon, no longer marketed in the US): capsules and, in the UK formulary, patches (MedlinePlus; NHS trust formulary).
• Galantamine (brand Razadyne, no longer marketed in the US): tablets, extended-release capsules and oral solution (MedlinePlus).
• Memantine (brand Namenda): an NMDA-receptor blocker as tablets, oral solution and extended-release capsules (MedlinePlus).
• Fixed combination: Namzaric (memantine extended release with donepezil 10 mg), for moderate to severe disease in people already stable on donepezil 10 mg (US label).
• Status: prescription-only. The NHS says only specialists such as psychiatrists or neurologists start cholinesterase inhibitors, and a GP may continue them on specialist advice (NHS).

How they work

Alzheimer's disease involves loss of acetylcholine-using brain cells. Cholinesterase inhibitors block the enzyme that breaks acetylcholine down, so more of it remains available. The Cochrane Lewy body review makes the same argument for dementia with Lewy bodies, where acetylcholine deficits are also found (Rolinski 2012). Galantamine is described as a specific, competitive, reversible inhibitor (Loy and Schneider 2006).

Memantine is a moderate-affinity, uncompetitive NMDA receptor antagonist (McShane 2019). MedlinePlus classes it as an NMDA receptor antagonist, and we did not find a human mechanism proof in the sources we opened. In humans what is proven is a modest improvement on test scores. These drugs do not remove the plaques or tangles, and no trial we opened shows that they slow the underlying disease.

What they are prescribed for

Licensed and guideline uses

• US donepezil label: "dementia of the Alzheimer's type", with efficacy shown in mild, moderate and severe disease (openFDA label).
• US memantine label: moderate to severe dementia of the Alzheimer's type (openFDA label).
• NICE TA217 / NG97 (England): donepezil, galantamine and rivastigmine are "options" for mild to moderate Alzheimer's; memantine alone for moderate disease if cholinesterase inhibitors are not tolerated or contraindicated, or for severe disease; for people already on a cholinesterase inhibitor, consider adding memantine in moderate disease and offer it in severe disease. Start with the lowest acquisition-cost inhibitor (NG97 text).
• Lewy body dementia (NG97): offer donepezil or rivastigmine in mild to moderate disease, consider them in severe disease, consider memantine if the inhibitors are not tolerated, and consider galantamine only if the others are not tolerated. At June 2018, donepezil, rivastigmine and memantine did not have a UK licence for this use (NG97 text).
• Vascular dementia: NG97 says to consider these drugs only if Alzheimer's, Parkinson's dementia or Lewy body dementia is also suspected, and not to offer them for cognitive impairment caused by multiple sclerosis (NG97 text).

The NICE site blocked automated access, so we read these recommendations in a December 2018 NHS trust formulary that reproduces the guideline text. The wording may have changed since; check the current guideline.

What works and what does not

Benefits by claim

Do they help memory and thinking?

Slightly, for most people who take them for several months. The donepezil Cochrane review found a 2.67-point ADAS-Cog advantage on a 70-point scale and a 1.05-point MMSE advantage on a 30-point scale at about 26 weeks (Birks 2018). For comparison, the DOMINO-AD team set the minimum clinically important difference at 1.4 SMMSE points (Howard 2012). The pooled MMSE gain is just under that bar. Rivastigmine (ADAS-Cog -1.79) and galantamine (-3.1 at 24 mg) are in the same range (Birks 2015; Loy and Schneider 2006). The NHS says the three work about equally well, and that response varies from person to person (NHS).

Do they keep people independent for longer?

This is where the independent trial parts company with the makers' trials. AD2000 randomised 565 community-living people with mild to moderate Alzheimer's. Over three years it found no significant difference in entering institutional care (relative risk 0.97, 95% CI 0.72 to 1.30) or in disability progression (relative risk for the combined outcome 0.96, 95% CI 0.74 to 1.24). Its authors concluded that donepezil's benefits fell below minimally relevant thresholds (Courtney et al. 2004). The later Cochrane review pooled more trials and found small gains in daily living but no difference in total healthcare resource use (Birks 2018). Both can be true: a small average gain on a rating scale need not translate into a different life course.

Does memantine add to donepezil?

In DOMINO-AD, adding memantine to donepezil gave no significant benefit over donepezil alone, while memantine itself improved SMMSE by 1.2 points (0.6 to 1.8) and BADLS by 1.5 points (0.3 to 2.8) versus placebo (Howard 2012). The Cochrane memantine review found concomitant cholinesterase inhibitors did not change the overall effect in moderate to severe disease (McShane 2019). Memantine was also less likely to be reported with agitation as an adverse event (RR 0.81, 0.66 to 0.99; moderate certainty), though it did not reduce agitation as a treatment effect.

What does stopping do? The deprescribing evidence

The main trial is DOMINO-AD. It took 295 community-living people with moderate to severe Alzheimer's (SMMSE 5 to 13) who had already taken donepezil for at least three months and randomised them to four groups for 52 weeks: continue donepezil, stop it, stop it and start memantine, or continue it and add memantine. Continuing donepezil beat stopping it by 1.9 SMMSE points (1.3 to 2.5) and 3.0 BADLS points (1.8 to 4.3), The cognitive gap exceeded the trial's own minimum clinically important difference (1.4 points); the function gap (3.0) was statistically significant but fell short of the 3.5-point threshold. The authors concluded that continuing donepezil had cognitive benefits above the minimum clinically important difference and significant functional benefits over 12 months (Howard 2012). NICE's guideline text says not to stop cholinesterase inhibitors in Alzheimer's because of disease severity alone (NG97 text). Limits: DOMINO-AD enrolled people who had tolerated donepezil, so it cannot say what happens to someone who is having troublesome side effects. We did not find tapering schedules in the sources we opened, and we give none.

Lewy body and Parkinson's disease dementia

The Cochrane review pooled six trials with 1,236 people, mostly in Parkinson's disease dementia, and found benefits on global rating, cognition, behaviour and daily living. Only one trial (McKeith 2000, rivastigmine) included Lewy body dementia, and the review found no MMSE benefit there. Adverse events were more common (OR 1.64, 1.26 to 2.15), as were dropouts (OR 1.94, 1.33 to 2.84), and tremor (OR 2.71, 1.44 to 5.09), although falls and motor scores did not differ significantly (Rolinski 2012). A later meta-analysis of ten trials reached a similar view: inhibitors improved MMSE, memantine did not, and rivastigmine carried more adverse events (RR 1.19) (Wang 2015). In Japan, Eisai's phase 2 donepezil trial (140 people, 12 weeks) reported MMSE gains of 3.8 points at 5 mg (2.3 to 5.3) and 2.4 at 10 mg (Mori 2012). Its confirmatory phase 3 trial (142 people) did not confirm superiority on its co-primary measures, although MMSE improved at 10 mg (2.2 vs 0.6, P=0.016) (Ikeda 2015). Because antipsychotics can worsen movement symptoms and cause severe sensitivity reactions in Lewy body dementia, NICE gives this group particular warnings (NG97 text).

Higher doses and different forms

The 23 mg donepezil tablet was no better than 10 mg in two studies, with more adverse events and withdrawals (Birks 2018). The label notes a 1,434-person trial where 23 mg beat 10 mg on one measure (SIB) but not on the clinician global rating (CIBIC-plus), and that 23 mg raised ulcer, bleeding and weight-loss rates (US label). The rivastigmine review suggests the patch may cause fewer side effects than capsules at similar efficacy (Birks 2015).

Risks and all side effects

Common side effects: donepezil

Pooled placebo-controlled trials in mild to moderate Alzheimer's (355 on placebo, 747 on donepezil 5 or 10 mg) (US label):

Across trials, donepezil users were more likely to report at least one adverse event (72% vs 65%; OR 1.59, 1.31 to 1.95) and to withdraw (24% vs 20%; OR 1.25, 1.05 to 1.50) (Birks 2018). MedlinePlus adds frequent urination, headache, dizziness, confusion, mood changes and hallucinations to the list of reported effects (MedlinePlus).

Common side effects: rivastigmine and galantamine

Rivastigmine users were about twice as likely to have an adverse event (OR 2.16, 1.82 to 2.57) or to withdraw (OR 2.01, 1.71 to 2.37) (Birks 2015). MedlinePlus lists nausea, vomiting, stomach pain, loss of appetite, weight loss, diarrhoea, tremor, sweating and sleep problems for rivastigmine, and similar effects plus depression for galantamine (rivastigmine; galantamine). The NHS says nausea, vomiting and loss of appetite usually ease after about two weeks (NHS).

Common side effects: memantine

Overall adverse events did not differ from placebo (RR 1.03, 1.00 to 1.06; high certainty). Dizziness was about 1.6 times as likely (6.1% vs 3.9%) and headache about 1.3 times (5.5% vs 4.3%). There was no difference in falls (McShane 2019). The US label lists dizziness, headache, confusion and constipation as the most common reactions, with discontinuation for adverse reactions of 10.1% on memantine vs 11.5% on placebo (openFDA label).

Serious and rarer risks

All interactions

Who should avoid them

Contraindications and cautions

• Allergy: the donepezil label contraindicates known hypersensitivity to donepezil or piperidine derivatives; memantine, to memantine or its excipients (US donepezil label; openFDA memantine label).
• Heart rhythm problems: tell the prescriber about a slow, fast or irregular heartbeat, heart block or previous fainting (MedlinePlus). The Beers criteria advise avoiding these drugs in older adults whose syncope may be caused by a slow heart rate (AGS Beers 2023).
• Stomach ulcers or bleeding, asthma, COPD, seizures, difficulty passing urine: labelled cautions for donepezil (US label).
• Kidney disease (memantine): for severe renal impairment (creatinine clearance 5 to 29 mL/min) the US label recommends a target of 5 mg twice daily (openFDA label).
• Multiple sclerosis cognitive impairment: NICE says not to offer these drugs (NG97 text).
• Mild cognitive impairment: not recommended for galantamine because of the death signal (Loy and Schneider).

Pregnancy and breastfeeding

Dementia drugs are rarely relevant in pregnancy, but the labels address it. For donepezil there are no adequate human data, the label says it "may cause fetal harm", and late-gestation dosing in rats increased stillbirths; there are no data on human milk (US label). The memantine label says use only if the benefit justifies the risk to the fetus (openFDA label).

Children

The labels and guidance we opened concern adults. We did not find paediatric use described and none is expected.

Older adults and falls

Nearly all users are older, so the fall and fracture signal matters. Memantine showed no difference in falls in the Cochrane review, whereas the Canadian cohort found higher hip fractures with cholinesterase inhibitors (McShane 2019; Gill 2009).

Dosage and how to take them

The prescriber sets the dose. The ranges below are label and formulary doses given for context only. They are not a guide to dosing.

Doses studied

Most donepezil trials used 5 or 10 mg a day, and two used a 23 mg slow-release form (Birks 2018). The memantine review analysed the licensed dose at 6 to 7 months (McShane 2019).

How long before it helps, and for how long

The main trials assessed benefit at 12 to 26 weeks (Birks 2018). The memantine reviewers said a longer trial is needed to see whether benefit lasts beyond six months (McShane 2019). NICE says to continue inhibitors regardless of severity alone, and current NHS advice is to continue them into later stages (NHS). The NHS page does not say when or how to stop; decisions to stop belong with the specialist and the family.

Practical points

Donepezil is taken in the evening just before bedtime, with or without food. Galantamine tablets are taken with food. Rivastigmine capsules are taken with meals. If rivastigmine is missed for more than three days, MedlinePlus says to check with the doctor before restarting because a lower dose is likely (donepezil; galantamine; rivastigmine). Generic products vary in appearance, so carers should check that the strength matches the prescription.

Follow the money: who makes them and who funded the evidence

Who made and sells them

• Donepezil: Eisai Inc. holds the original US applications for Aricept: NDA 020690 (5 and 10 mg, approved 25 November 1996) and NDA 022568 (23 mg, 2010), as listed by openFDA (openFDA). Eisai Co., Ltd. is headquartered in Tokyo, Japan, according to the sponsor statement in its Lewy body trial (Mori 2012 full text). The Cochrane review lists Pfizer support for several donepezil trials and joint Pfizer-Eisai support for others, which fits a shared development and marketing relationship; we did not open a source describing the commercial agreement itself (Birks 2018 full text).
• Memantine: the Namenda application, NDA 021487 (approved 16 October 2003), lists AbbVie as sponsor and the brand as discontinued. Namzaric, NDA 206439, was approved 23 December 2014 under AbbVie (openFDA). A 2006 Cochrane conflict statement calls Forest Pharmaceuticals a memantine maker, and the Cochrane memantine review mentions Lundbeck-funded studies (Loy and Schneider; McShane). We did not trace the chain of ownership between Forest and AbbVie.
• Rivastigmine and galantamine: the Cochrane records point to Novartis and Janssen as the makers' trial sponsors and the brands are no longer marketed in the US (Rolinski; Loy and Schneider; MedlinePlus).
• Generics: openFDA lists 21 donepezil and 29 memantine generic (ANDA) applications (donepezil; memantine). NICE advises starting with the lowest acquisition-cost inhibitor, which usually means a generic (NG97 text). Little commercial incentive remains to fund new trials of the old molecules.

Who funded the evidence

• Pivotal donepezil trials: the Cochrane review's source list shows Eisai Inc. supporting Rogers 1996, Rogers 1998a and b, Burns 1999 and Tariot 2001; Pfizer supporting Winblad 2001 and 2006 and Black 2007; and Eisai China sponsoring Jia 2017. Overall, 17 of 30 trials were industry-funded and 4 independent (Birks 2018 full text). The Cochrane team also contacted the Donepezil Study Group and Eisai Inc. for data (abstract).
• Does funding bend the result? Across 14 trials, industry-funded trials had a pooled SMD of 0.46 (0.38 to 0.54) against 0.33 (0.18 to 0.48) for independent ones, a gap that did not reach statistical significance. The authors say the non-significant result may reflect chance or low statistical power (Killin 2014). That is a disclosure, not a verdict: the independent trials still show a benefit.
• Public trials: AD2000 is listed by Cochrane as funded by NHS Executive R&D (West Midlands), and Cochrane says it ran independently of the pharmaceutical company. DOMINO-AD was funded by the UK Medical Research Council and the Alzheimer's Society, with drug and matching placebo supplied by Eisai and Lundbeck (DOMINO-AD protocol). Drug supply by a maker is a tie worth knowing even though the trial was publicly funded.
• Lewy body trials: Mori 2012 and Ikeda 2015 were sponsored by Eisai Co., Ltd.; the second says the sponsor helped with design, data collection and analysis, and lists two Eisai employees as authors (Mori; Ikeda). All the rivastigmine trials with usable data were industry-funded or sponsored (Birks 2015).
• Reviewer ties: the memantine review's lead author declared investigator-initiated studies funded by Lundbeck; the galantamine review's senior author declared contracts and consulting links with several makers; the Lewy body review's author declared past conference support from Eisai, Shire and Novartis (McShane; Loy and Schneider; Rolinski).

Documented integrity events

We searched for regulator actions, lawsuits or recalls tied specifically to these medicines but found none in the sources we opened. We make no claim either way. The documented evidence-quality concerns are the sponsor-linked trial results above and the galantamine death signal in mild cognitive impairment.

What this means: The pooled evidence that these drugs improve test scores rests heavily on trials paid for by their makers, but it is not contradicted by the independent work. The publicly funded AD2000 and DOMINO-AD trials, and Cochrane reviews by authors without industry ties for donepezil, find real but small effects. The independent evidence disagrees with the makers' framing only on how much the effect matters in people's lives. With the drugs now largely generic, the unanswered questions (long-term benefit, safe stopping, mild-disease memantine) are unlikely to be funded by industry.

Related research

For wider, evidence-graded context, see these Pure City Research guides:

• Bradycardia: slow heart rate, normal variation, diagnosis and treatment
• Heart block: AV conduction problems, degrees, symptoms and treatment
• Vasovagal syncope: reflex fainting, assessment, prevention and treatment
• Orthostatic hypotension: blood-pressure drops on standing and care
• REM sleep behavior disorder: assessment, treatments and safety
• Sleep disorders in older adults: ageing, insomnia, breathing and safe care
• Depression in older adults: assessment, treatment evidence and safety
• Sleep and neurological disease
• Stroke: emergency signs, types, treatment and recovery

Frequently asked questions

Sources and funding notes

What we could not verify: (1) The NICE website blocked automated access, so we read NG97 and TA217 recommendations in an NHS trust's December 2018 formulary (text may have been updated since). (2) AD2000's full text is paywalled; its funder comes from the Cochrane review (NHS Executive R&D, West Midlands), while the PubMed record carries a US National Institute on Aging grant number we could not reconcile, and we could not confirm whether Eisai or Pfizer supplied the drug. (3) We saw only the abstract of DOMINO-AD and the protocol for its funding and drug supply. (4) Funding statements were not in the records we read for Gill 2009, Wang 2015, McKeith 2000 or the Cochrane rivastigmine and Lewy body reviews. (5) The Beers Criteria disclosure section was not in the part we read. (6) We did not research prices, market shares, the Pfizer-Eisai agreement, or the chain of ownership from Forest to AbbVie, and we found no regulator action specific to these medicines. (7) A 2024 update of the Cochrane galantamine review exists and we did not read it. (8) The DOMINO-AD full text and AD2000 critiques were not reviewed.

1. Birks JS, Harvey RJ. Donepezil for dementia due to Alzheimer's disease. Cochrane 2018 (full text) — UK; role: main donepezil review; authors report no known interests; 17 of 30 trials industry-funded.
2. McShane R et al. Memantine for dementia. Cochrane 2019 — UK; role: memantine review; UK Department of Health grant 13/89/22; one author's Lundbeck-funded investigator studies disclosed.
3. Birks JS, Grimley Evans J. Rivastigmine for Alzheimer's disease. Cochrane 2015 — UK; role: rivastigmine review; no funding statement in the record; included usable trials industry-funded.
4. Loy C, Schneider L. Galantamine for Alzheimer's disease and mild cognitive impairment. Cochrane 2006 — Australia/US; role: galantamine review; authors' Janssen, Pfizer, Novartis and Forest ties disclosed.
5. Rolinski M et al. Cholinesterase inhibitors for dementia with Lewy bodies, Parkinson's disease dementia and cognitive impairment in Parkinson's disease. Cochrane 2012 — UK; role: Lewy body and Parkinson's dementia review; one author's Eisai, Shire and Novartis support disclosed.
6. Wang HF et al. Efficacy and safety of cholinesterase inhibitors and memantine in Parkinson's disease, Parkinson's disease dementia and dementia with Lewy bodies. J Neurol Neurosurg Psychiatry 2015 — country not verified; role: meta-analysis; funding not in the record read.
7. Courtney C et al. Long-term donepezil treatment in 565 patients with Alzheimer's disease (AD2000). Lancet 2004 — UK; role: publicly funded trial; NHS Executive R&D (West Midlands) per Cochrane.
8. Howard R et al. Donepezil and memantine for moderate-to-severe Alzheimer's disease (DOMINO-AD). NEJM 2012 — UK; role: deprescribing trial; Medical Research Council grants G0600989, MC_U122888466, MC_U137686861 and the Alzheimer's Society.
9. DOMINO-AD protocol (PMC) — UK; role: funding and drug-supply statement (Eisai and Lundbeck supplied medication and placebo; MRC and Alzheimer's Society funding).
10. Killin LOJ et al. The effect of funding sources on donepezil randomised controlled trial outcome. BMJ Open 2014 (full text) — UK; role: funding-bias meta-analysis; no specific grant.
11. Gill SS et al. Syncope and its consequences in patients with dementia receiving cholinesterase inhibitors. Arch Intern Med 2009 — Canada; role: safety cohort; funding not stated in the record.
12. McKeith I et al. Efficacy of rivastigmine in dementia with Lewy bodies. Lancet 2000 — UK/Spain/Italy; role: single DLB trial; funder not stated in the record read.
13. Mori E et al. Donepezil for dementia with Lewy bodies: a randomized, placebo-controlled trial. Ann Neurol 2012 (full text) — Japan; role: DLB phase 2; sponsored by Eisai Co., Ltd.
14. Ikeda M et al. Donepezil for dementia with Lewy bodies: a randomized, placebo-controlled, confirmatory phase III trial. Alzheimers Res Ther 2015 (full text) — Japan; role: DLB phase 3; sponsored by Eisai Co., Ltd., two Eisai employee authors.
15. Paulison B et al. Potential cardiotoxic reaction involving rivastigmine and beta-blockers: a case report and review of the literature. Cardiovasc Toxicol 2010 — record title only; role: interaction case report.
16. American Geriatrics Society 2023 Beers Criteria — US professional society; role: avoid-in-syncope and anticholinergic-in-dementia entries; disclosures not seen.
17. NICE TA217 and NG97, read via an NHS trust formulary (December 2018) — UK; role: guidance and UK doses.
18. NHS: Alzheimer's disease treatment — UK NHS.
19. openFDA: Aricept label; openFDA: Namzaric label; openFDA: memantine oral solution label — US government data (unvalidated, per openFDA); manufacturer-supplied label text.
20. openFDA: donepezil NDAs; memantine NDAs; ANDA counts for donepezil and memantine — US government application records.
21. MedlinePlus: donepezil (revised 15 August 2026), memantine (15 July 2026), rivastigmine (15 August 2026), galantamine (15 August 2026) — US National Library of Medicine.

Educational research reviewed 6 October 2026. This article is not medical advice; talk to a qualified clinician or pharmacist before starting, stopping or combining treatments.