The results of the AD2000 study of donepezil (Aricept) in Alzheimer's—published in the June 26 Lancet—made a big splash in the media pond. While confirming previous findings of modest benefits on certain tests, the study authors challenge claims that the drug has valuable effects on "real-life" measures of disability, and they charge that is not cost-effective.

The use of cholinesterase inhibitors (ChEIs) has exploded in just a few years, producing golden eggs for several pharmaceutical companies. Detractors have noted that the benefits are far from outstanding. In particular, there is no strong clinical trial evidence to suggest that patients should stay on the drugs for more than a year or two, though many apparently do (see ARF Live Discussion on ChEIs). While many short-term studies indicate beneficial effects on neuropsychological tests and activities of daily living scales, there has been little study of whether the drugs provide benefits over the long term, especially in terms of overall disability and the need for institutionalization.

The publicly funded AD2000 study, conducted by the Clinical Trials Unit of the University of Birmingham in England, was created to fill this void, and also to address concerns that the research to date has mainly been industry-funded, with design biases toward showing benefits. AD2000 was intended to assess donepezil in a "real-clinic" situation, in contrast to industry trials, whose inclusion and exclusion criteria tend to exclude the majority of dementia patients. Originally slated to enroll 3,000 patients for a single phase of 60 weeks, the study ended up with only 566. The study was redesigned to follow all these patients through a series of phases.

The blame for the low turnout can be laid on a variety of forces, argue the authors, including the rapid speed with which ChEIs became available to patients at large. Once this occurred, it became difficult to recruit patients for a trial in which they might be receiving a placebo. This lower number of subjects means the study lost statistical power to make conclusive assessments about its primary outcomes of institutionalization and progression of disability, writes Lon Schneider of the University of Southern California in Los Angeles in an editorial accompanying the study.

Even into the trial, these same forces exerted a strong influence—the high attrition rate is partly due to patients opting for open-label treatment with ChEIs. Of the 486 patients entering the 48-week Phase 1, only 293 completed that phase; 111 patients completed Phase 2; and 20 patients completed phase 3. However, other factors such as death and institutionalization also contributed to the drop in subjects.

Despite the attrition rate, the study did not lose power for the secondary endpoints; these include changes in specific tests scores such as MMSE or the Bristol activities of daily living score (BADLS). Indeed, AD2000 confirmed and extended previous results, showing that in the donepezil subjects cognition averaged 0.8 MMSE points better, and functionality measured at 1.0 BADLS points better, and that this benefit was maintained over the first two years of the trial.

In terms of the primary endpoints, the AD2000 researchers report no differences between the drug and placebo groups in terms of institutionalization—42 percent vs. 44 percent at three years (p=0.4). Similarly, they report no differences in disability progression between donepezil and placebo groups—58 percent versus 59 percent at three years (p=0.4). The researchers also report finding no differences between donepezil and placebo in terms of behavioural and psychological symptoms, caregiver psychopathology, formal care costs, unpaid caregiver time, adverse events or deaths, or between 5 mg and 10 mg donepezil.

Despite the lack of power for primary endpoints, both the authors and commentator Schneider feel that the results deserve notice. The study results, writes Schneider, "…undermine an assumption that improvements in cognition and ADL scores generalize to maintenance of function, cost savings, or delay in institutionalization. … Results are incompatible with many drug-company-sponsored observational studies and advertisements claiming remarkable effects for cholinesterase inhibitors. For example, claims that donepezil stabilises cognitive decline, or delays nursing-home placement by 2-5 years now can be seen as implausible in the light of AD2000."

The authors argue that "AD2000 found no evidence that costs of caring for patients with Alzheimer's disease in the community are reduced by donepezil. Indeed, our findings suggest that these drugs would increase [British National Health Service] cost per patient—mostly hospital care—by around 40 percent without any offsetting through reductions in service use."

Some previous economic models have suggested that ChEI-related improvement on tests such as the MMSE will lead to cost reductions, but the authors argue that the studies these models were built upon "clearly overestimate benefit because of selection bias." The only study that, in the opinion of the authors, was properly set up to assess economic endpoints failed to find a benefit from donepezil (Wimo et al., 2003).

But wouldn't the small, but significant, MMSE and BADLS improvements with donepezil over the two-year study mean that patients experienced an increased quality of life? Without question, it is a difficult thing to assess, but the authors note that the U.S. FDA's guidelines on a worthwhile clinical response would require at least a 1.4 MMSE point difference, and many clinicians believe that the minimum clinical benchmark is a 3.0 point difference on the MMSE, both figures outside the confidence intervals for these data.

The authors make another argument against the interpretation of previous clinical trials of ChEIs. "The significantly worse cognitive and functional decline seen among retrieved dropouts in AD2000 compared to those who continued treatment confirms that differential dropout, particularly if analyses do not include retrieved dropouts, exaggerates efficacy," the authors write, noting that differential dropout was not discussed in previous clinical trial reports.

What about separating responders from non-responders early on and sparing expenses and pointless side effects for the non-responders? The potential for this appears to be low. "Reliable discrimination between responders and non-responders—on the basis of changes on simple scale outcome measures of cognition, functionality, and behavioural and psychological symptoms—is not possible," argue authors.

Ultimately, it is those who pay for the drugs—whether public or private—who will make the decisions about their availability. In the popular press, medical statistician Robert Gray, director of the University of Birmingham Clinical Trials Unit and spokesman for the study, has argued that in England the money spent on cholinesterase inhibitors would be better spent on doctors and nurses and better social support for patients and their caregivers.—Hakon Heimer


  1. The AD2000 study reported in The Lancet is an important article, and Lon Schneider's editorial also is important. The findings were that "donepezil produces small improvements in cognition and activities of daily living in patients with mild to moderate Alzheimer's disease" with no loss of benefit for 2 years. It was not cost-effective when measured in terms of institutionalization or progression of disability. Thus, the benefits are modest at best.

    I do not believe that any clinician managing Alzheimer's patients believes differently: donepezil (indeed, all the cholinesterase inhibitor drugs approved for the treatment of Alzheimer's disease) show consistent but modest benefit in virtually all the studies performed to date. The question is whether the modest benefit is "worthwhile". As the AD2000 authors point out, it is difficult to evaluate what is "worthwhile". I believe that many patients and their families appreciate even the modest benefits these drugs afford. Of course to some extent this is supported by the sense of hope that comes from being able to treat this terrible disease with drugs that are only minimally beneficial versus the despair of having no treatment options at all.

    There is no question that better drugs are needed to treat the symptoms of Alzheimer's disease. More importantly, drugs are needed that target the disease mechanisms of Alzheimer's disease; these drugs may then arrest progression of the illness or, if introduced early enough in the incipient stages of the disease, conceivably even prevent the appearance of symptoms. Although I will continue to offer cholinesterase inhibitor drugs to my patients with mild-moderate Alzheimer's disease, we all should support the development of truly effective therapies.

  2. The AD 2000 collaborative group have studied the cost-effectiveness of donepezil in the UK. In their study, donepezil did not delay nursing home placement, and since this is the major cost driver of Alzheimer's disease (AD), the agent did not reduce overall care costs of AD.

    In considering this study, it is important to note that it verified small but significant effects of donepezil on cognition and activities of daily living (ADL) and showed that the effect was evident over the two years of the trial—longer than any previous trial. From a patient perspective, these are very important observations and do not undermine the use of cholinesterase inhibitors in practice. This should be separated clearly from the issue of cost effectiveness, i.e., a payer perspective.

    Several aspects of the study deserve comment. First, in an effort to make the study as relevant to actual clinical practice as possible, relatively brief diagnostic assessments and outcomes measures were applied. This resulted in a high frequency of co-morbidity and likely inclusion of some non-AD types of dementia. Second, patients were included if the doctor was "substantially uncertain that the individual would obtain a worthwhile clinical benefit from donepezil"; that is, the physicians were searching for non-responders to enter into the trial. There is no information on exactly how these decisions were made but the methods seem to have been successful. This approach doesn’t simulate clinical practice, where the doctor does not try to eliminate those likely to respond to treatment but instead include them. Third, three past studies in the U.S. Knopman et al., 1996; Lopez et al., 2002; Geldmacher et al., 2003) have suggested that treatment with cholinesterase inhibitors delayed nursing home placement.

    None of these studies is without flaws, but it is uncertain whether it is differences in the studies, differences between the health care systems of the UK and the U.S., (which are substantial), or other differences that account for these varying outcomes. The results of the study by the AD 2000 collaborative group are not necessarily generalizable to other systems.

    The AD 2000 collaborative group has performed a valuable service in attempting to develop a methodology for generating effectiveness information that differs from the efficacy information derived from clinical trials. These results, however, should not be interpreted as demonstrating that donepezil and other cholinesterase inhibitors are not worthwhile. They provide a different type of information relevant to the specifics of this trial, in a particular setting, and most applicable to health finance issues in the UK.


    . Long-term tacrine (Cognex) treatment: effects on nursing home placement and mortality, Tacrine Study Group. Neurology. 1996 Jul;47(1):166-77. PubMed.

    . Cholinesterase inhibitor treatment alters the natural history of Alzheimer's disease. J Neurol Neurosurg Psychiatry. 2002 Mar;72(3):310-4. PubMed.

    . Donepezil is associated with delayed nursing home placement in patients with Alzheimer's disease. J Am Geriatr Soc. 2003 Jul;51(7):937-44. PubMed.

  3. This study could not guarantee correct typing of dementias. As such, non-AD cases could have been included.

  4. David S. Geldmacher, MD “AD2000” is a long-term, double-blind, placebo-controlled study of donepezil in patients with Alzheimer’s disease. In the results published in the Lancet paper, the authors found that donepezil treatment resulted in higher cognitive test scores and improved scores on an assessment of daily function in comparison to placebo. They also assessed a number of other outcomes including nursing home placement, caregiver burden, and frequency of progression in functional disability, but found no evidence that donepezil provided benefit in these outcomes. The authors then concluded, “Donepezil is not cost effective.”

    That definitive statement about cost-effectiveness has drawn significant media attention, and raised further controversy in the already contentious discussions about the value of symptomatic therapy in AD. It has also provoked major criticism from the Alzheimer’s advocacy community and many researchers who study treatments for AD.

    How can the results of this study differ so much from previous papers about AChEI therapy and outcomes like reduced risk for, and delayed time to, nursing home placement? As with most things in science, the devil is in the details.

    Previous studies reporting the positive outcomes have been structured differently, conducted differently, and used different populations from the AD2000 study. Each factor contributes to the differences between AD2000 and the prior reports. AD2000 was a double-blind placebo controlled study. This is usually considered the gold standard for treatment studies, and therefore might have been weighed more heavily than the retrospective case-matching or open-label follow-up methods of previous studies. (e.g., Lopez et al., 2002; Geldmacher et al., 2003). However, several shortcomings prevent the AD2000 study from being definitive.

    First, it was originally intended to enroll 3,000 patients, but only 565 actually entered, of whom 482 entered long-term follow-up. If the original design required 3,000 patients, then 565 would be insufficient to provide a definitive answer to the questions. The authors provide questionable statistical arguments that 565 patients were sufficient. But, if that is true, then it is hard to explain why 3,000 was the original target. Either 3,000 were too many (and therefore scientifically unjustified or unethical) or 565 were too few (and therefore scientifically incomplete). The argument that 565 were probably too few is most strongly supported by the fact that only 4 patients out of the 565 entered Phase 4 of the study.

    Second, the population was more clinically diverse than past clinical trials. This study allowed patients with cerebral vascular disease (strokes) to enroll, as long as the clinician judged that the dementia was at least partially due to AD. While this more closely reflects typical clinical practices, it will contribute to different observations than a group of people with “pure” AD. Also, cognitive and physical deficits attributable to stroke were not separated in the results. Past studies (e.g., Black et al., 2002) have shown that there is less response to AChEIs in patients with dementia attributable to cerebral vascular disease than among those with AD.

    Perhaps the most important difference from previous trials was the study entry requirement that the physician should be “substantially uncertain” that treatment would benefit the patient. Therefore, the trial excluded any patient considered by the physician to be likely to respond to treatment well. This process likely eliminated the best responders and biased the study away from finding treatment effects. In contrast, differential retention in the drug and placebo groups may have influenced drug-placebo differences. Patients and families whose needs were being met well on placebo were less likely to discontinue the study and move over to open label treatment. Therefore, the placebo group, especially later in the study, probably performed better than would be the case by simple random chance.

    In summary, the AD2000 trial was a laudable attempt to answer the question of “How effective is donepezil (and by extension, the other cholinesterase inhibitors) for the treatment of Alzheimer’s disease?” However, practical limitations in enrollment, patient selection, and retention prevent the study from being conclusive.

    Certainly, the statement “Donepezil is not cost-effective” is not warranted by the published results. The authors did not find evidence for cost-effectiveness in a study less than one-sixth of the size they had determined to be necessary to answer the question. The “absence of proof” in this study is by no means a “proof of absence” of the effectiveness of cholinesterase inhibitors.

    How should the results of AD2000 affect clinical decision making for AChEI therapy? The bulk of evidence in the literature, and a finding supported by AD2000, is that AChEIs provide a small average benefit for patients with mild and moderately severe AD in domains like cognition, daily function, and behavior, in comparison to placebo. Also, a proportion of AD patients (perhaps one in five) will show an obvious short-term improvement in one or more of those domains. Therefore, an individualized therapeutic trial to see if a patient improves or stabilizes seems well supported by the existing literature and is not refuted by AD2000.

    Improvement on the Mini-Mental State Exam should not be the sole criterion for judging treatment responsiveness. Clinical stabilization is valuable to patients and caregivers, and may involve other domains like behavior, daily function, or perceived caregiver burden (Winblad et al., 2001). If a patient’s decline continues unabated on AChEI treatment after an adequate trial (perhaps three months), then discontinuation might be considered. AD2000 suggests that continued treatment of non-responders is not useful, but the methodological weaknesses preclude a definitive interpretation.

    If treatment is discontinued, the prescriber should closely monitor for the adverse effects of discontinuation. Prior studies have shown that donepezil discontinuation can have a negative long-term impact (e.g., Doody et al., 2001). Therefore, a patient who shows easily discerned loss of cognition or other function in the first week off therapy was probably a treatment responder, and therapy should be resumed as soon as possible to prevent irreversible loss of ability. The AD2000 study did not incorporate the impact of treatment discontinuation into its decision structure, to the possible detriment of its enrollees. The protocol-mandated treatment discontinuations in the AD2000 donepezil group also reduce the validity of the study’s drug-placebo comparisons.

    The question of how to assess whether a person is deriving relative benefit from treatment is important and remains problematic. Certainly, if the MMSE is stable and the caregiver is reporting no behavioral change or functional loss in an AD patient, then the treatment goal of stabilization is being met. If the person is clearly worsening over six-month intervals, then it is likely that treatment is no longer providing maximal benefit and a trial discontinuation may warranted. In 2004, many physicians will consider the addition of memantine to AChEI therapy rather than AChEI discontinuation.

    There is an opportunity for society to benefit from the conflict over differences between the AD2000 results and prior studies. The energy from this controversy may be able to drive further research on more effective ways of studying these complex questions. Ultimately, the goal of such research should be to provide the clinician with more useful information on whom to treat, how to measure response, and when to stop therapy in AD. Better research will be necessary for the field of AD therapeutics to strike a meaningful balance between a person’s desire for optimized well-being and the public’s needs for real value in the treatment of the rapidly nexpanding AD population.


    . Efficacy and tolerability of donepezil in vascular dementia: positive results of a 24-week, multicenter, international, randomized, placebo-controlled clinical trial. Stroke. 2003 Oct;34(10):2323-30. PubMed.

    . Open-label, multicenter, phase 3 extension study of the safety and efficacy of donepezil in patients with Alzheimer disease. Arch Neurol. 2001 Mar;58(3):427-33. PubMed.

    . Donepezil is associated with delayed nursing home placement in patients with Alzheimer's disease. J Am Geriatr Soc. 2003 Jul;51(7):937-44. PubMed.

    . Cholinesterase inhibitor treatment alters the natural history of Alzheimer's disease. J Neurol Neurosurg Psychiatry. 2002 Mar;72(3):310-4. PubMed.

    . Pharmacotherapy of Alzheimer's disease: is there a need to redefine treatment success?. Int J Geriatr Psychiatry. 2001 Jul;16(7):653-66. PubMed.

  5. My simplified view is that the study was unable to answer the main questions that it posed, so it is largely uninformative with respect to effectiveness outcomes. That having been said, the study at least suggests that the effectiveness of these agents is modest at best, consistent with what all other studies have shown. The study also suggests the possibility of no effectiveness, or even "anti-effectiveness," but it is inconclusive, as Lon Schneider said well in his editorial.

    The problem I see with some of the lay media coverage on this issue has been the exaggeration of the confidence in the findings. Another point is that we do not treat averages, we treat individuals. These results will not influence my practice. They do underscore the need for proper effectiveness studies so that we can speak from evidence and not conjecture.

  6. I thought Aricept gave my father a painful bilateral arthritic reaction of both hands and both feet, which led to his being treated briefly with steroids (which, incidentally, brought a discernible temporary improvement in his mentation) and at length with methotrexate. What a joke for someone to take a drug whose beneficial effects are fairly negligible in order to end up on a toxic drug requiring blood tests for liver function etc. Now he's still on methtrexate (long off Aricept) because no one wants to go to the trouble of stopping the methotrexate to see if the arthritis will come back.

    The rheumatologist, when I inquired as to whether he thought the artritis could have come from the Aricept, was arrogant and dismissive of the very idea, although it is listed as a possible side effect in the PDR. I did also try to find out from Aricept's manufacturer whether they knew much about this and/or had other people reporting it, but they were certainly dismissive and seemed rather condescending instead of being helpful.

    I think in this patient population something should have a very robust therapeutic effect before we subject patients and their families to it, and especially when there may be a down side. In the absence of a robust therapeutic effect, the main effect we can count on is profits for the manufacturers. I do not know any psychiatrists who think this drug is really worth giving but they cave in to pressure from patients and family members, such that it functions as a sort of placebo by proxy.

    Laura Fisher

  7. ChEIs Under Fire: An Issue Too Complex for Sound Bites
    Sam Gandy

    In April 2004, the front page of The New York Times (in prime position above the fold) carried the headline, “Minimal Benefit Seen for Alzheimer’s Drugs”. Adding fuel to the fire, in June of this year a second piece in the same paper summarized the results of the “AD2000” study, published in Lancet, under the headline “British Study Sees Scant Value in…Aricept”. Is this really news? How should professional and family caregivers react?

    Personally, my standard line, which appeared in The Times and has been widely reprinted, is that cholinesterase inhibitors (ChEIs) provide “modest, temporary benefits for some people." Nothing in these recent reports or the ensuing debate has led me to revise this language. Indeed, the academic clinical Alzheimerologists whose opinions I have sought expressed no surprise at all. The typical reaction of these specialists to the AD2000 study has uniformly been that “AD2000 confirmed what we already knew.” The difference now is that there exist hard data on the failure of ChEIs to delay institutionalization or to diminish costs of care. This, in turn, can influence 1) demand by consumers, 2) prescription activity by physicians, and 3) qualification for reimbursement by third-party payers (i.e., health care insurers such as PPOs, HMOs, and Medicare).

    The AD2000 study was complex in design: 565 patients entered the trial, which lasted two to four years. Subjects were initially randomized to drug or placebo for 3 months, then treatment was halted for a “washout,” after which time 486 were re-randomized to drug or placebo for a year (48 weeks). After a second washout, 293 subjects went on to the second year. 111 completed the second year. Two further washout/treatment cycles were performed, by which time only a handful of patients remained participants.

    The small number of subjects completing the trial to the very end of four years has been used to criticize the results; however, 111 did complete the second year, by which time the trends in the data appear to have been established. The expected, typical initial benefit in cognitive function with ChEI use was observed, excluding the possibility that AD2000 subjects represented some variant, ChEI-resistant Alzheimer’s. Yet this early improvement did not translate into measurable impact on outcomes of either cost-benefit or time to nursing home admission. One novel result was that washout and re-institution of therapy showed that drug response did not suffer from washout/re-treatment, as has been feared in clinical practice. Allaying this fear is important since current practice frequently involves permanent therapy from diagnosis to death, for fear that an irreversible step-decline would be revealed. The new data will provide reassurance that ChEIs can be withdrawn and, if diminished function results, the drugs can be re-instituted with the expectation that the pre-washout functional level can be re-attained.

    What is the “take-home message”? Are these widely prescribed medications useless? Of course, this depends on how you define “useful”. Clearly, since the argument has gone on unresolved for years, there will be no obvious, unequivocal, “wow!” benefit. The U.S. FDA is satisfied that ChEIs are useful (at least sometime, in some patients). Imagine a scenario in which cognitive function improves in someone suffering from early Alzheimer to the point that he or she recognizes loved ones longer or is able to put his or her affairs in order. This could certainly be inestimably valuable, but the benefits would be quite limited in scale and unlikely to show up on cost-benefit analysis.

    Probably the more typical situation is that of the desperate caregiver in the clinician’s office following the initial visit at diagnosis. These are the circumstances under which ChEI therapy is frequently initiated. Failure to offer a ChEI, the only medication approved for this indication, would serve to reinforce and deepen the desperation on the parts of all concerned. Sending the family home with a prescription for a ChEI serves to buck up their spirits as this horrible news sets in. Fortunately, ChEIs are safe and non-toxic, but is this an appropriate use of an expensive prescription medication?

    What if ChEIs were suddenly withdrawn or disallowed for third-party reimbursement? Where would the tragedy strike? ChEI programs in pharma now provide the major funding stream for support of Alzheimer drug discovery worldwide, the fruits of which have yielded the first generation of anti-amyloid therapeutics that are entering the clinic today. Pharma investment in Alzheimer drug discovery exceeds academic support worldwide by severalfold. In an ironic twist of real-world economics, abandonment of ChEIs might drastically weaken, or even arrest, the development of the pharma drug discovery programs that, for the first time in the history of medicine, target molecular pathogenesis.

    On balance, I am a self-professed proponent of ChEI use but, in my heart of hearts, I know that my position is based, at least in part, on reasoning that has little to do with demonstrable benefit or any of the usual criteria for approval and support of drugs.

    See also:
    Grady, D. “Minimal benefit is seen in drugs for Alzheimer’s” (April 7, 2004) and “British study shows scant value in Alzheimer drug Aricept” (June 25, 2004). The New York Times.

    Official comment from the Alzheimer’s Association.


    . Long-term donepezil treatment in 565 patients with Alzheimer's disease (AD2000): randomised double-blind trial. Lancet. 2004 Jun 26;363(9427):2105-15. PubMed.

  8. It's sad that pharma's hype about Alzheimer drugs leads families that cannot afford these top shelf drugs to feel so guilty. Patients know that a drug exists for their loved one and believe it would be negligent to withhold it. The industry preys upon the dire needs of AD victims by charging them outrageous prices for drugs that do little or nothing. Even if a percentage of patients gets a tiny bit better, that in no way dents the tragedy of the disease, salvages the person's dignity, or improves prognosis. These drugs are the emperor's new clothes. They not only let pharma reap billions, but they assuage the physicians' sense of helplessness before his or her patients. Most physicians base their judgement of drug efficacy on a simple enquiry of a family member as to whether they see improvement. On the basis of an anecdotal remark the physician is all too ready to place a significant financial burden on the family.

    Even worse is the common occurrence when the drug is having no observable effect, but the physician keeps the patient on the medication according to the purely hypothetical unsubstantiated rationale that the patient would be doing worse off drug. The blind trust put in us by many of our patients translates into gratitude, and hence a double deception that the patient has a remedy and the doctor has a happy patient. How can such a small increment in an ADAS-Cog score justify such a furious marketing campaign? The minuscule benefits do not justify the costs for many patients. I can easily think of a gizillion better ways to spend money on drugs which do have an impact—like delivery of AIDS meds to the third world. Alzheimer meds are an example of our distorted priorities.—A concerned neurologist

  9. The AD2000 study was evidently intended to answer the question of whether or not donepezil is worth its cost. Before looking at the study, we should examine its context. "Worth" was defined here primarily as improvement in symptoms or quality of life, or delayed institutionalization. The latter endpoint in particular is influenced by many social factors above and beyond response to therapy. The questioners are healthcare providers, not patients and caregivers. I would maintain that stabilization of symptoms or slowing of decline is much more important to patients and families than actual improvements. Further, in the U.K., the National Health Service has been historically reluctant to pay for AD treatments of any type. As this study was being planned and initiated, another large-scale study was completed that influenced the NHS to allow wider treatment (NICE), a decision that was resisted by many healthcare providers. My concern regarding the framing of the question in AD2000 is that this study was designed to show that donepezil is not worth the cost.

    Many people have commented on the methodological issues of this complicated study design. The authors initially sought to reduce and control bias whenever possible. They sought large numbers of subjects in a naturalistic setting, and utilized rigorous analytic techniques. Unfortunately, they failed at some of these aims, but went ahead with the analysis. First, they failed to recruit the 3,000 subjects they intended and settled for 566. Why? Because the National Health Service (NHS) made the drug available on a limited basis during the recruitment phase, and many patients sought treatment. Those who did not and agreed to be randomized must differ from those who entered the study. The design of this study required long-term follow-up since the value of an AD treatment must be assessed over the course of the disease, but 177 of 282 patients randomized to drug and 150 of 283 randomized to placebo left the study for reasons unrelated to the treatment, either because they withdrew (usually to go on treatment) or stopped therapy (quite possibly for the same reason). Many would argue that the study is invalid since it considers only those who remained, at sites where no one withdrew. These are most likely sites at which the investigators did not value treatment.

    My biggest concern with the design has to do with the dosing and the several required washout periods built into the study. After years of treating patients, I am convinced that working patients up to the maximally tolerated dose of a cholinesterase inhibitor and persisting in therapy without interruption are the most important principles to successful treatment. In this study, the first randomization involved 12 weeks at the starting dose of 5mg/day or placebo. Patients were then re-randomized, so that persistent treatment was not guaranteed. After each 48-week period, they were washed out for 4-6 weeks (long enough to lose benefits). The study therefore failed to examine the benefits of optimal treatment and persistent therapy, which is the question that we would all like to know the answer to. The huge dropout rate throughout the study compromises the analysis of each phase.

    There are many specific criticisms that I have not addressed, such as the definition of “loss” of ADL’s, the significance of a 0.3 SD effect size, etc. but my comment is already quite long. In a way, this study illustrates what not to do in the long-term treatment of AD patients. It should not claim to have proven that “Donepezil is not cost effective, with benefits below minimally relevant thresholds.”

  10. Back to the Roots: The Original Lancet Paper
    The somewhat turbulent debate about the AD2000 paper highlights the need to go back to the basic source, i.e., the original Lancet paper, and to do so with a particular focus on the methods used and the results presented.

    The role of cholinesterase inhibitors in the treatment of AD has always been under debate, particularly in terms of cost-effectiveness and the clinical relevance of the clearly significant effects on cognition shown in 3- to12-month studies.

    People with AD may live 10 years or more with their disease (although the average survival of AD populations is less than that). Clinical trials have a much shorter duration, making it difficult to catch the long-term effects. Therefore, several model approaches have been used. The basic idea is that models with different techniques extrapolate short-term results (often in terms of cognition or severity) to longer periods. The role of models is always under debate since models per definition are not “empirical,” and drug authorities, evidence-based medicine groups, and many clinicians are sceptical about models. For that reason, long-term empirical studies are preferable.

    Thus, the Lancet paper is indeed welcome, particularly since the approach is a randomized, controlled trial with an intention-to-treat, retrieved drop-out approach. However, with an approach like this, it is inevitable that there will be methodological problems, particularly regarding drop-outs. Even if the different types of drop-outs were similar between the groups, their magnitude is nevertheless a problem. Mortality and institutionalization are of course natural endpoints in a study like this, but the total number of withdrawn patients was 37 percent in both groups after the second randomization. Since the original recruitment plan aimed for 3,000 patients and the final study population consisted of 566 patients before the first randomization and 482 in the second randomization, there are problems with statistical power. The withdrawal problem in this trial is similar to the problems we faced in the Nordic donepezil trial (Winblad et al., 2001), where Aricept was launched in Sweden during the recruitment period.

    In the public debate, the AD2000 study is discussed as a three-year trial, which is only true for the primary endpoints. For the scale results, it is a two-year trial, and for the health economics evaluation, data for 60 weeks are presented (the latter is based on all phases of the trial, making the “n” high).

    The inclusion criteria aimed at being more representative of a general AD population than previous studies. Perhaps it is, but, as in other trials, a regular carer is demanded. This excluded many AD patients, particularly with mild AD, who are living in the community without carers. Such AD patients can only be found in population-based studies.

    The analysis of institutionalization is unclear. The number of patients at risk at baseline was 282 and 283, respectively (fig 2), which is the number of patients before the first randomization according to the flow-chart in fig 1. Since there was a second randomization after the run-in period, the figure probably does not illustrate the same groups of patients (after the run-in period, patients were again randomized). Even if the run-in period is short, and the results hardly would change if the analysis focused on the patients after the second randomization, I think this way of presentation is odd (or perhaps I have misunderstood it?).

    The rejection of the cost-neutrality hypothesis is unclear to me. There is no summary table for all costs from a societal perspective with all the costs (including drug costs and costs of informal care) with 95 percent confidence intervals. The costs in table 2 include formal care costs. The argument that costs of informal care not are included in the cost calculations since the differences were not significant is doubtful and could consequently be used on all cost components. In our health economics analysis of the Nordic donepezil trial (Wimo et al., 2003), we presented one-year data including the cost of drug and informal care, resulting in a non-significant difference. Of course, it is not as easy to say that a non-significant difference supports cost neutrality since a low power (with a wide confidence interval) could give such false support, but I think this should be up to the reader to judge. The health economics calculations in both these trials are probably underpowered.

    In conclusion, AD2000 is nevertheless an important trial, as Lon Schneider says in his comment in Lancet. Even if it has methodological problems and fell short by far of its recruitment goal, it is a large study. It shows no long-term effects in postponing institutionalization or preserving activities of daily living. It confirms previously shown results on cognition, now extended to two years. The rejection of cost-neutrality is unclear, at least for me. Whether treatment with donepezil can be regarded as cost-effective or not is indeed debatable, but the AD2000 results are not as dramatic as presented in the media. Nevertheless, further long-term studies are needed.

    As a clinician and researcher, I have during the years had a rather pragmatic view of the cholinesterase inhibitors. They are better than nothing but the effects are modest, and they are expensive.


    . A 1-year, randomized, placebo-controlled study of donepezil in patients with mild to moderate AD. Neurology. 2001 Aug 14;57(3):489-95. PubMed.

    . An economic evaluation of donepezil in mild to moderate Alzheimer's disease: results of a 1-year, double-blind, randomized trial. Dement Geriatr Cogn Disord. 2003;15(1):44-54. PubMed.

  11. As a fairly inexperienced clinician working with dementia diagnostics and treatment, I'd appreciate more comments by the experts on the AD2000 protocol in regard to all those quite long washout periods, the first one as early as after 12 weeks. In my training it has been stressed that it is essential to keep up continuous medication lest the deteriation will accelerate to quickly reach a stage comparable to that of untreated patients. Might it be the case that this repeated washout protocol in a way "ensured" the mediocre findings? Was it wished for? Perhaps it also contributed to the extensive dropping out - if patients and their carers actually feared discontinuity of medication? If the study had been funded by the pharmacological industry, would they have allowed such washout periods? With what rationale were they necessary for this study?

    View all comments by Olle Kjellin
  12. I agree wholeheartedly with Dr. Kjellin. I was very surprised when I reviewed the AD2000 protocol closely and realized that the study design allowed frequent and long washout periods and in the statistical design did not analyze the donepezil 5 mg and 10 mg treatments groups separately but lumped both groups together during analysis. This is equivalent to concluding that a diabetes drug does not work to lower hemoglobinAIC when the dose is below therapeutic levels and dose is withheld for several weeks. The AD2000 study had the potential to be groundbreaking in that it was the first, non-industry funded, prospective trial. However, it appears that the investigators designed the study to "stack the deck" against seeing any therapeutic benefit from cholinesterase inhibitors. What a shame. There will likely not be an opportunity in the near future to have this type of study, and for good reason, enough benefit HAS been proven with cholinesterase inhibitors that placebo-controlled trials of CHeI are unethical.

    View all comments by Diana Kerwin

Make a Comment

To make a comment you must login or register.


Paper Citations

  1. . An economic evaluation of donepezil in mild to moderate Alzheimer's disease: results of a 1-year, double-blind, randomized trial. Dement Geriatr Cogn Disord. 2003;15(1):44-54. PubMed.

Other Citations

  1. ARF Live Discussion

Further Reading

No Available Further Reading

Primary Papers

  1. . Long-term donepezil treatment in 565 patients with Alzheimer's disease (AD2000): randomised double-blind trial. Lancet. 2004 Jun 26;363(9427):2105-15. PubMed.
  2. . AD2000: donepezil in Alzheimer's disease. Lancet. 2004 Jun 26;363(9427):2100-1. PubMed.