Published on 16 September 2026
Donepezil and reimbursement: what a Cochrane review can tell a family, and what it cannot
In brief
This Cochrane review brings together 30 double-blind randomised trials and 8257 participants with dementia due to Alzheimer’s disease, 28 of which could be meta-analysed. The main analysis, which compares donepezil 10 mg/day with placebo at 24 or 26 weeks, rests on 13 trials and 3396 participants. At this dose, donepezil does better than placebo on cognition measured with the ADAS-Cog, a scale running from 0 to 70 points: mean difference −2.67 (95% CI −3.31 to −2.02; 1130 participants, 5 trials). On the MMSE, the difference is 1.05 points (95% CI 0.73 to 1.37; 1757 participants, 7 trials). On the SIB battery, which runs from 0 to 100, it is 5.92 points (95% CI 4.53 to 7.31; 1348 participants, 5 trials). The benefit is also present on activities of daily living, measured in patients at the severe stage, and on the clinician-rated global impression of change. There is, on the other hand, no demonstrated difference on behavioural symptoms or on quality of life, two dimensions that weigh heavily in the daily lives of families. Against this, 72% of patients on donepezil report at least one adverse event against 65% on placebo (OR 1.59; 95% CI 1.31 to 1.95; 2500 participants, 10 trials), and 24% stop treatment before the end of the trial against 20% (OR 1.25; 95% CI 1.05 to 1.50; 2846 participants, 12 trials). The authors conclude that there is moderate-quality evidence of benefits they describe as small. In France, the drug remains authorised and can be prescribed, but it has not been reimbursed since 1 August 2018. The example is French, the structure of the problem is not.
The context
A review from 2018, and exactly what that means
Let us say it at the outset: this review is eight years old, and its search window is more than nine. Its search closed on 20 May 2017, it was published on 18 June 2018, and no later version had been posted at the time of writing. That is how Cochrane works: a review is up to date at the date of its search, not at the date it is read. In practical terms, everything published on donepezil since May 2017 lies outside the scope of the figures quoted here, and readers need to bear that in mind before reusing them.
A Cochrane review that has not been updated is not a wrong review. Its method remains valid and its calculations remain correct. Its corpus, however, is frozen: whatever may have appeared after May 2017 is not in it. We did not run an updated search for this analysis, and we therefore do not claim that nothing has appeared since. What we do say is that donepezil is a drug authorised in France since 1997, whose commercial development is complete, and that the probability of a large placebo-controlled trial having come along to overturn this picture is low without being nil. That is an appraisal, not a datum.
The point that makes the difference in consultation
A relative arrives with a simple and legitimate question: why is this medicine no longer covered, when it can still be obtained from the pharmacy? Two notions then become confused, even among well-informed relatives, and the confusion feeds a sense of abandonment. A marketing authorisation attests that the benefit-risk balance is judged favourable for a given indication. Reimbursement is a separate decision, which answers a different question: does this benefit justify coverage by the community? Actual marketing, in between, depends on the company holding the authorisation, and it is a third question again. In France, donepezil is the textbook case of authorisation and coverage coming apart. The example is French, the structure of the problem is not: wherever medicines are funded collectively, authorisation, marketing and coverage are distinct decisions, and a decision to stop reimbursing a drug is not in itself evidence that the drug does not work.
The French example, checked item by item
| Question | Verified answer | Source consulted on 10 August 2026 |
|---|---|---|
| Is donepezil still authorised in France? | Yes. Marketing authorisation granted on 3 September 1997 for the originator 10 mg product, status marketed | French public medicines database (ANSM, HAS, statutory health insurance) |
| For what indication? | The French product information indicates it for the symptomatic treatment of Alzheimer’s disease in its mild to moderately severe forms | Package leaflet and summary of product characteristics, French public medicines database |
| Is it reimbursed in France? | No. The product record states that the price is set freely and that the medicine is not reimbursable, and specifies that the presentation is not listed for use by public bodies | French public medicines database |
| Since when exactly? | Delisting effective from 1 August 2018, ordered by two orders of 29 May 2018 published in Journal officiel no. 0124 of 1 June 2018: one for the list of medicines reimbursable to insured persons, the other for the list of medicines admitted for use by public bodies | Légifrance, JORFTEXT000036970192 and JORFTEXT000036970203 |
| On what grounds? | Opinion of the Transparency Committee of the Haute Autorité de santé of 19 October 2016, which concluded that the medical benefit of the originator donepezil products was insufficient to justify their coverage by national solidarity | Haute Autorité de santé, donepezil record |
| Do French guidelines still mention it? | Yes. The care pathway guide of the Haute Autorité de santé (May 2018) cites the four drugs, recalls the 2016 opinion on insufficient medical benefit and states that, since they retain a marketing authorisation, these drugs may be prescribed | Haute Autorité de santé, care pathway guide for patients with a neurocognitive disorder |
Three practical consequences follow from this table, in the French example. The drug can be prescribed and dispensed: nothing has been withdrawn from the market, and the French guidelines continue to discuss it. It is paid for by the patient, at a freely set price, which should be announced before the first dispensing and not discovered at the pharmacy counter. And the question is directly a psychiatric one, not only a neurological or geriatric one: wherever psychiatrists initiate or renew this treatment, the conversation with the family falls to them.
A final clarification on what the French decision is not. Insufficient medical benefit is a French regulatory category, built on the severity of the condition, efficacy, safety, place in the treatment strategy and public health interest. It is a legal category, not a trial result. The Cochrane review analysed here was never intended to say whether reimbursement was justified, and it does not say so. The same holds outside France: a coverage decision weighs criteria that go beyond the size of an effect, and it cannot be read as a verdict on efficacy.
The study at a glance
| Question (PICO) | |
|---|---|
| Population | |
| People with dementia due to Alzheimer’s disease, mild to severe forms. 30 studies, 8257 participants, 28 of which reported their results in enough detail to be meta-analysed. Of the 30 studies, participants had mild to moderate disease in 21 studies, moderate to severe in 5 and severe in 4. The main analysis, donepezil 10 mg/day versus placebo at 24 or 26 weeks, brings together 13 studies and 3396 participants; 11 of these studies were multicentre, 7 recruited patients at the mild to moderate stage, 2 at the moderate to severe stage and 4 at the severe stage, with a mean age of about 75 years | |
| Intervention | |
| Oral donepezil, mainly 5 mg/day or 10 mg/day. Two studies compared 10 mg/day with 23 mg/day, delivered by a slow-release oral formulation | |
| Comparator | |
| Placebo, in parallel groups. The review also analyses comparisons between doses of donepezil | |
| Outcomes | |
| Cognitive function, activities of daily living, behavioural symptoms, global clinical state, quality of life, adverse events, deaths, healthcare resource use and costs | |
| Design | |
| Cochrane systematic review with meta-analysis, restricted to double-blind randomised trials of at least 12 weeks. Update of a review first published in 1998. Search closed on 20 May 2017, covering the specialised register of the Cochrane Dementia and Cognitive Improvement group, MEDLINE, Embase, PsycINFO and other sources; the authors state that they also contacted members of the Donepezil Study Group and the company Eisai Inc. Certainty graded with GRADE, outcome by outcome · CEBM 1a |
Quality control
| Criterion | Status |
|---|---|
| Level of the synthesis | Sound |
| FindingCochrane review restricted to double-blind randomised trials, GRADE assessment outcome by outcome, results reported with the number of studies and participants entering each estimate | |
| Data extraction | Reservation |
| FindingThe authors write: “One reviewer (JSB) extracted data on cognitive function, activities of daily living, behavioural symptoms, global clinical state, quality of life, adverse events, deaths and healthcare resource costs.” Independent double extraction is nonetheless the usual rule. The point deserves to be flagged, even when the rest of the work is careful | |
| Currency of the corpus | Reservation |
| FindingSearch closed on 20 May 2017, publication in June 2018, no update since. The corpus is frozen, which invalidates nothing but dates everything | |
| Length of the trials | Reservation |
| FindingMost studies last six months or less. Only one small study reaches 52 weeks. A disease that progresses over years is not judged over six months | |
| Funding of the included trials | Reservation |
| Finding17 studies funded or sponsored by industry, 4 independent, 9 with no information on the source of funding. The synthesis is independent, its raw material much less so | |
| Authors’ competing interests | Reservation |
| FindingThe PubMed record of the review carries a conflict of interest statement for both authors: “Jacqueline Birks: none known” and “Richard Harvey: none known”. It is a declaration, not an independent check, and we draw no further conclusion from it in either direction. The contact with the manufacturer of donepezil, mentioned in the methods, reflects a common practice of retrieving unpublished data and does not in itself constitute a conflict of interest | |
| Grading of certainty | Reservation |
| FindingModerate GRADE certainty for almost all outcomes, downgraded for risk of bias: insufficient information on allocation concealment and on blinding of outcome assessment. Moderate means that the true estimate is probably close to the observed estimate, but that it could differ substantially from it | |
| Fit between claim and evidence | Sound |
| FindingThe authors themselves speak of “small benefits in cognitive function, activities of daily living and clinician-rated global clinical state”, report the absence of difference on behaviour and quality of life, and document the increase in adverse events. The conclusion does not go beyond the data | |
| Scope relative to the French indication | Reservation |
| FindingThe review covers mild, moderate and severe forms. The SIB data come from the most advanced forms, at the limit of or beyond the French wording of the indication, which covers mild to moderately severe forms | |
The findings
Unless otherwise stated, all the results below compare donepezil 10 mg/day with placebo at 24 or 26 weeks, and almost all carry moderate GRADE certainty.
| Outcome | Published result |
|---|---|
| Cognition (ADAS-Cog, 0 to 70) | Mean difference −2.67 (95% CI −3.31 to −2.02); 1130 participants, 5 studies |
| PEB readingEstablished and small the interval clearly excludes zero. The next question, the only useful one, is whether 2.67 points out of 70 can be seen in daily life | |
| Cognition (MMSE, 0 to 30) | Mean difference 1.05 (95% CI 0.73 to 1.37); 1757 participants, 7 studies |
| PEB readingOne MMSE point the order of magnitude is the same on the scale best known to clinicians and families | |
| Cognition (SIB, 0 to 100) | Mean difference 5.92 (95% CI 4.53 to 7.31); 1348 participants, 5 studies |
| PEB readingAdvanced stages the relative gap is a little more marked, but the corresponding population extends beyond the wording of the French indication | |
| Activities of daily living (ADCS-ADL, version for severe disease) | Mean difference 1.03 (95% CI 0.21 to 1.85); 733 participants, 3 studies |
| PEB readingLower bound close to zero the result is positive, on three studies only, and the interval remains compatible with a very small effect | |
| Global clinical state (clinician-rated impression of change) | Odds ratio 1.92 (95% CI 1.54 to 2.39); 1674 participants, 6 studies |
| PEB readingClear difference but a composite outcome rated by an assessor, and therefore sensitive to any partial unblinding | |
| Behavioural symptoms (NPI) | Mean difference −1.62 (95% CI −3.43 to 0.19); 1035 participants, 4 studies |
| PEB readingNo difference demonstrated the interval crosses zero. It remains compatible with a modest benefit as well as with none at all | |
| Behavioural symptoms (BEHAVE-AD) | Mean difference 0.4 (95% CI −1.28 to 2.08); 194 participants, 1 study |
| PEB readingInconclusive a single study, a small sample, a wide interval | |
| Quality of life | Mean difference −2.79 (95% CI −8.15 to 2.56); 815 participants, 2 studies |
| PEB readingInconclusive the interval is very wide. This is not a demonstration of absence of effect, it is an absence of demonstration | |
| Withdrawal before the end of treatment | 24% versus 20%; odds ratio 1.25 (95% CI 1.05 to 1.50); 2846 participants, 12 studies |
| PEB readingUnfavourable one patient in four did not finish, under trial conditions more closely supervised than real life | |
| Adverse events (at least one, at 26 weeks) | 72% versus 65%; odds ratio 1.59 (95% CI 1.31 to 1.95); 2500 participants, 10 studies |
| PEB readingUnfavourable seven more patients per hundred treated report at least one adverse event | |
| Healthcare resources and costs | The authors find no evidence of a difference between donepezil and placebo in total healthcare resource use |
| PEB readingHandle with caution no difference shown does not mean equivalence demonstrated. The result was not produced in the French health system and settles nothing about whether reimbursement is warranted | |
The question of dose
A point of method first, because it is often glossed over. The table of figures above, with its GRADE assessment, concerns the comparison of donepezil 10 mg/day with placebo. The question of dose is not addressed by a second comparison of 5 mg/day with placebo carrying the same level of grading, but by direct comparisons between doses, based on few studies and reported without numerical estimates in the review abstract. Statements about dose therefore rank lower in the hierarchy of evidence than statements about efficacy against placebo, and they should be read as such.
Three studies compared 10 mg/day with 5 mg/day over 26 weeks. The 5 mg dose was associated with slightly worse cognition on the ADAS-Cog, but not on the MMSE or the SIB, with slightly better quality of life, and with fewer adverse events and treatment withdrawals. Two studies compared 10 mg/day with 23 mg/day: no difference on efficacy outcomes, and fewer adverse events and withdrawals on 10 mg/day. Summing this up as 10 mg doing better than 5 mg is an unwarranted shortcut: the gap appears on only one cognitive scale out of three, and it is paid for in tolerability as well as in quality of life. What these data allow us to say is more modest: going above 10 mg/day finds no justification in these trials, and the choice between 5 and 10 mg/day is a trade-off between an uncertain cognitive gain and poorer tolerability.
What 2.67 points mean, and what they do not
Three clarifications, in order of their clinical usefulness.
First, the order of magnitude. The ADAS-Cog has 70 points, the MMSE 30, the SIB 100. The observed differences represent about 3.8%, 3.5% and 5.9% of the range of each scale respectively. This calculation is a simple division, it says nothing about what the patient perceives, but it sets a frame: we are not talking about a change of trajectory.
Next comes the question of thresholds, and this is where precision is needed, because precise values circulate without any identifiable source. The Cochrane review itself defines no threshold of clinical relevance for the ADAS-Cog: it reports mean differences, not threshold crossings. There is indeed one published estimate, that of Schrag and Schott in 2012, obtained by an anchor-based method in 181 patients at the mild stage from the ADNI cohort: patients whose clinician judged their worsening clinically significant had lost on average 3.1 to 3.8 ADAS-Cog points, and distribution-based approaches gave similar values, 3.2 points for half a standard deviation and 3.7 for the standard error of measurement. The authors concluded that a worsening of 3 points could constitute a minimal clinically relevant change in trials at the early stage.
Three reservations rule out applying this figure as it stands to the result of the review. It was built on worsening, not on improvement, and the authors themselves point out that the improvement side remains to be documented. It applies to mild disease, whereas the review mixes stages. And it is an isolated estimate, presented as such, not a validated and consensual threshold. There is therefore, to date, no established threshold of clinical relevance for the ADAS-Cog that can be relied on. In its absence, the honest comparison remains that of the amplitude against the range of the scale, with the admission that its translation into lived experience is not established. We will simply note that 2.67 points lies below the only published reference value we have been able to trace.
Last, the very nature of the measure. A difference in means between two groups describes no patient in particular. It is compatible with a situation in which a few patients derive a clear benefit and the majority derive none, just as with a situation in which everyone gains a little. Group data do not allow a choice between these two worlds, and that is precisely what one would like to know before prescribing.
Critical appraisal
| Domain | Judgement |
|---|---|
| Robustness of the design | Sound |
| FindingThirty double-blind randomised trials, GRADE assessment, conclusions calibrated to what the data allow. The level of evidence of the synthesis is the highest there is for this drug | |
| Observation horizon | Reservation |
| FindingSix months for most of the corpus, a single small study at one year. Long-term benefit is neither demonstrated nor refuted, it is not tested | |
| Outcomes that matter to families | Reservation |
| FindingBehaviour and quality of life are the two dimensions on which nothing is demonstrated, and they are the ones that determine whether the patient can stay at home and whether carers become exhausted. Care is needed, however, not to turn the argument around: the intervals remain wide, particularly for quality of life (from −8.15 to 2.56 across two studies) | |
| Balance of benefit and tolerability | Reservation |
| FindingThe benefit is small and the excess of adverse events is clear. That is not prohibitive, it is a trade-off to be set out explicitly with the patient and those close to them rather than decided on their behalf | |
| External validity | Reservation |
| FindingA trial population, with a mean age of about 75 years in the main analysis, and comorbidity and polypharmacy less present than in the consulting room. Real patients are older, frailer and on more medication | |
| Transposition to a coverage decision | Reservation |
| FindingThis review sheds light on the size of the effect, it does not rule on reimbursement. In the French example, medical benefit is a matter of French law and draws on elements other than effect sizes alone. Confusing the two makes Cochrane say what Cochrane does not say | |
| Extraction and currency | Reservation |
| FindingData extracted by a single review author, search closed in May 2017. Two limitations of different kinds, one methodological, the other temporal, both of which call for caution without disqualifying the work | |
It is worth separating the three registers explicitly. What is demonstrated, with moderate certainty: over six months, donepezil produces a small benefit on cognition, on activities of daily living and on the clinician’s global impression, and it increases adverse events and treatment withdrawals. What is not demonstrated: an effect on behavioural symptoms, an effect on quality of life, a benefit beyond one year. On these three points, the reading is not demonstrated, not demonstrated to be absent, and the nuance is not one of style. To this should be added the absence of any difference shown in total healthcare resource use, which belongs to the same register and is no demonstration of economic equivalence. What is suggested, on comparisons between doses based on two or three studies and without their own grading: that 23 mg/day adds nothing to 10 mg/day, that tolerability worsens as the dose rises, and that 5 mg/day does slightly less well on only one of the three cognitive scales while doing slightly better on quality of life. What is opinion, our own included: whether a gap of this size does or does not justify collective coverage, and what a clinician does with a small benefit when there is nothing else to offer.
Level of evidence
PEB appraisal: moderate confidence in the existence of a small cognitive and functional benefit at six months, moderate confidence in the increase in adverse events and treatment withdrawals, no conclusion possible on behaviour, quality of life or outcome beyond one year. What lowers the scientific score has nothing to do with the conduct of the review, which is rigorous and honest about its limitations, but with three elements: data extraction by a single author, the near-universal brevity of the included trials and the weight of industry funding in the primary corpus. The editorial score reflects the immediate usefulness of this analysis in consultation, tempered by the fact that the scientific message itself is not new.
The colleague test
What an experienced colleague would say if you put this analysis to them in two minutes, between two consultations.
“ The benefit is real, it is small, and it bears on scores, not on what the family comes in to describe to me. On agitation and on quality of life, I have nothing to announce. Against that, one patient in four stops before six months. So I can prescribe it, I have to say what it costs, and I have to set in advance the date on which I reassess. ”
What this means in practice: the right words in front of a family are neither that it is useless nor that it works, they are by how much, on what, and at what price.
What you can do with this on Monday morning. Where the drug is authorised but no longer covered, as in France, prescribing remains possible. It is the conversation that needs preparing.
- Separate the decisions clearly in front of the family: authorisation, marketing and collective coverage. In the French example, donepezil is still authorised, for the symptomatic treatment of mild to moderately severe forms, it is still marketed and it can still be prescribed. It has not been reimbursed since 1 August 2018, under two orders of 29 May 2018, themselves based on an opinion of the Haute Autorité de santé of October 2016 concluding that the medical benefit was insufficient. No authority said that the drug was dangerous or that it did nothing: it was judged not useful enough to be paid for by the community. A reimbursement decision is not proof of ineffectiveness, wherever it is taken.
- Give orders of magnitude rather than an opinion. Over six months, at 10 mg/day, as a group average, about 2.7 points on an ADAS-Cog scale that has 70, about 1 MMSE point. Nothing demonstrated on behavioural symptoms or on quality of life. About 7 more patients per 100 report at least one adverse event, and 4 more per 100 stop before the end. These figures can be given in a minute and they are worth more than a position.
- Announce the cost before the first dispensing where the drug is not covered. In the French example, the price is set freely and the medicine is not reimbursable, including in hospital, where the presentation is not listed for use by public bodies. A family that discovers the amount at the pharmacy counter lives through a second piece of bad news, and an avoidable one.
- If you start or continue treatment, write down the stopping rule before you begin. Set the target dose, the reassessment date and the criterion that will trigger stopping. A reminder on dose is useful: do not exceed 10 mg/day, since tolerability worsens as the dose rises, and do not take it for granted that 10 mg/day is better than 5 mg/day, since the gap appears on only one cognitive scale out of three and is paid for in adverse events and quality of life. Watch gastrointestinal tolerability in particular: the review cites nausea, vomiting and diarrhoea as the most frequent adverse events, most often mild in intensity.
- Do not let the end of reimbursement decide for you, in either direction. After August 2018, in a cohort of 19,380 patients followed in French memory centres, 19.5% stopped treatment, with increased use of psychotropic drugs, antidepressants in particular, and more non-drug interventions, without a significant difference in MMSE scores the following year. This is a before-and-after observation, without a control group, in a memory centre population: it describes a reality on the ground, it demonstrates no causality. Above all, it is a reminder that stopping a treatment is never neutral and that it should be prepared as a decision, not suffered as an administrative consequence.
- Reinvest consultation time in what no reimbursement decision has ever touched: assessing carer exhaustion, detecting depression and sleep disorders, correcting sensory deficits, anticipating advance directives, coordinating with the memory centre. Remember, finally, that wherever psychiatrists initiate or renew this treatment, the question is as much theirs as the neurologist’s or the geriatrician’s: that makes you legitimate on this matter, including when the answer is no.
Frequently asked questions
Does a drug that is no longer reimbursed lose its marketing authorisation?
No, the two are separate. In the French example, the marketing authorisation for donepezil remains in force and the drug is marketed, for the symptomatic treatment of Alzheimer’s disease in its mild to moderately severe forms. What changed on 1 August 2018 was reimbursement, not authorisation. Authorisation and coverage are two distinct decisions, taken by different authorities, on different criteria. The Cochrane review itself documents no regulatory or reimbursement status in any country.
Is a Cochrane review from 2018 still usable?
It remains the reference synthesis on this drug, but its corpus stops at 20 May 2017 and it has not been updated. Its method and calculations are not out of date, its body of literature is. We did not run an updated search for this analysis and we therefore do not claim that nothing has appeared since. A reader who has to decide today would do well to check this point before relying on these figures.
Does a statistically significant result mean a noticeable benefit?
No. Significance indicates that the observed gap is hardly compatible with chance, it says nothing about its size. With more than a thousand participants, a small gap becomes detectable. The next question is one of amplitude: 2.67 points on a scale that has 70, one MMSE point, can be measured without necessarily being visible in an ordinary day.
What should I tell a family who asks why donepezil is no longer reimbursed?
In the French example, that the Haute Autorité de santé concluded in 2016 that the medical benefit was insufficient to justify coverage by national solidarity, and that the government drew the consequences through two orders of 29 May 2018, taking effect on 1 August 2018. Then give the figures: a real but small benefit on cognitive scores at six months, nothing demonstrated on behaviour or quality of life, more adverse events. It is not abandonment, it is a trade-off that can be explained, and discussed.
Does the Cochrane review on donepezil justify ending its reimbursement?
No, and saying so would be a misreading. The review measures effects, it does not take a position on reimbursement policy. It finds no evidence of a difference between donepezil and placebo in total healthcare resource use, which is not the same thing as demonstrating equivalent cost, and this result was not produced in the French system. The two lines of reasoning intersect, they do not merge.
Should donepezil be stopped in a patient who has been taking it for years?
The review does not answer this question. It covers trials of at least twelve weeks comparing donepezil with placebo or with another dose, most of them over six months or less; stopping a long-term treatment is not the question it asks, and its results do not allow it to be answered. The decision is made case by case, on tolerability, on the benefit perceived by those close to the patient and on the financial burden borne by the patient where the drug is not covered, and it deserves to be taken explicitly rather than endured.
Which donepezil dose should be used if treatment is continued?
The answer is less clear-cut than is often said. In three studies at 26 weeks, 5 mg/day did slightly less well than 10 mg/day on the ADAS-Cog, but not on the MMSE or the SIB, and it did slightly better on quality of life, with fewer adverse events and fewer withdrawals. In two studies, 23 mg/day brought no gain in efficacy and caused more adverse events and withdrawals than 10 mg/day. What can be inferred without forcing it: going beyond 10 mg/day has no justification, and the choice between 5 and 10 mg/day is to be discussed case by case rather than settled by default. These comparisons rest on few studies and do not carry the same grading as the results against placebo.
Annotated bibliography
Birks JS, Harvey RJ (2018). Donepezil for dementia due to Alzheimer’s disease. Cochrane Database of Systematic Reviews, 6(6), CD001190. DOI 10.1002/14651858.CD001190.pub3 · PMID 29923184. The source study analysed here. Update of a review first published in 1998, literature search closed on 20 May 2017. Of the 30 included trials, 17 were funded or sponsored by industry, 4 were independent, and 9 had no information on the source of funding. The authors state that they contacted the Donepezil Study Group and Eisai Inc. as part of their search, a usual practice for gathering unpublished data. The PubMed record gives the conflict of interest statement as “Jacqueline Birks: none known” and “Richard Harvey: none known”; we draw no further conclusion from it. Full text in open access, PMCID PMC6513124; the complete version of the review was consulted for this analysis, including the summary of findings table and its GRADE assessment.
Schrag A, Schott JM (2012). What is the clinically relevant change on the ADAS-Cog? Journal of Neurology, Neurosurgery and Psychiatry, 83(2), 171-173. DOI 10.1136/jnnp-2011-300881 · PMID 22019547. The only estimate of a minimal clinically relevant change on the ADAS-Cog that we have been able to trace back to a primary source. Anchor-based method in 181 patients at the mild stage from the ADNI cohort, mean baseline score 18.5; mean worsening of 3.1 to 3.8 points in patients judged clinically worse, with concordant values from distribution-based approaches (3.2 points for half a standard deviation, 3.7 for the standard error of measurement). The authors propose a worsening of 3 points as a reference for trials at the early stage, and point out that the improvement side remains to be established. This is not a validated and consensual threshold, and it applies neither to improvement nor to advanced stages.
Haute Autorité de santé. Care pathway guide for patients with a neurocognitive disorder associated with Alzheimer’s disease or a related disease, May 2018. has-sante.fr. Source for the statement that the French guidelines still cite these drugs: the guide recalls the 2016 opinion on insufficient medical benefit, then states that, since they retain a marketing authorisation, these drugs may be prescribed. French regulatory context, cited here as an example.
Haute Autorité de santé. Donepezil, acetylcholinesterase inhibitor: opinion of the Transparency Committee of 19 October 2016. has-sante.fr. Conclusion: the medical benefit of the originator donepezil products is insufficient to justify their coverage by national solidarity. A French regulatory opinion, based on a set of criteria that goes beyond effect size alone.
Order of 29 May 2018 delisting pharmaceutical products from the list referred to in the first paragraph of Article L. 162-17 of the French Social Security Code. Journal officiel no. 0124 of 1 June 2018. legifrance.gouv.fr. Article 1: the products listed in the annex are removed, from 1 August 2018, from the list of medicines reimbursable to insured persons. The annex includes products based on donepezil, rivastigmine, galantamine and memantine.
Order of 29 May 2018 delisting pharmaceutical products from the list of medicines admitted for use by public bodies provided for in Article L. 5123-2 of the French Public Health Code. Journal officiel no. 0124 of 1 June 2018. legifrance.gouv.fr. Twin of the previous text, with the same effective date, which explains why the presentation is not listed for use by public bodies either.
French public medicines database. Originator donepezil 10 mg film-coated tablet, product record and package leaflet. base-donnees-publique.medicaments.gouv.fr. Marketing authorisation of 3 September 1997, valid status, marketed product, price set freely and medicine not reimbursable, presentation not listed for use by public bodies. Official source for the French regulatory status cited in this article, consulted on 10 August 2026.
Herr M, Ankri J, Diard C, Hiance-Delahaye A (2021). Removal of drugs for Alzheimer’s disease from the list of reimbursable drugs in France: analysis of change in drug use, disease management and cognition using the National Alzheimer Data Bank (BNA). Drugs & Aging, 38(1), 63-74. DOI 10.1007/s40266-020-00817-3 · PMID 33410119. Source of the real-world data cited in the practice box: 19,380 patients from the French National Alzheimer Data Bank, 62.5% women, mean age 81 years, 19.5% stopping treatment after the end of reimbursement, with increased use of psychotropic drugs and of non-drug interventions and no significant difference in MMSE scores the following year. A before-and-after observational study, without a control group, restricted to patients followed in memory centres, and therefore not representative of everyone concerned. It describes a change, it does not establish a cause-and-effect relationship.