When Aricept (donepezil) was first approved by the FDA in 1996 for the treatment of Alzheimer’s Dementia (AD), there was an almost euphoric sense that we were finally beginning to make progress in treating a previously untreatable disease.

While the data were not overwhelming, it was clear that patients on Aricept experienced a slower cognitive decline than those on placebo. Two other companies quickly rolled out their own cholinesterase inhibitors (CIs), both sporting data comparable to the original, though there were some significant differences in side effect profiles and ease of use, with Exelon conspicuously disadvantaged by a 31% incidence of vomiting as a side effect (see TCR Vol. 1, No. 6, 2003, for a review of Aricept, Reminyl – recently renamed “Razadyne” – and Exelon).

It was nice to finally have something effective to offer patients with AD.

Well, the situation is very different today. A recent influential article has thrown into question the whole notion that CIs are helpful in AD, and Great Britain is seriously considering a proposal to drop all CIs from the formulary of the National Health Service.

What’s going on here?

Let’s take a close look at the Lancet article that stirred things up (Lancet 2004; 363:2105-2115). Called “AD2000,” the study was conducted in Great Britain, funded entirely by the National Health Service, and was unique in that it measured the effectiveness of a CI over a three year period, rather than the standard six month length of most industry trials.

A total of 486 patients with AD were randomized to one of three arms: Aricept 5 mg QD, Aricept 10 mg QD, or placebo. In order to ensure that recruited subjects were representative of the populations most doctors are likely to see, patients were admitted if their clinicians believed they had AD according to DSM-IV criteria. There were no exclusion criteria – patients were admitted even if they had significant medical, neurological, or psychiatric comorbidity. In contrast, in most CI drug trials, inclusion requires symptom ratings with multiple research scales, neuro-imaging, and the lack of significant comorbidity. Such study criteria have been shown to exclude over 90% of community-diagnosed AD patients from studies, leading many to question the generalizability of such data to the patients we see in our offices (J Am Geriatr Soc 1997; 45: 923- 928).

At any rate, the primary outcomes variables in this study were: 1. Time to institutionalization; and 2. Time to disability (defined as the loss of the ability to perform certain basic activities of daily living). While this choice of outcome measures may seem reasonable, almost all industry-funded studies use very different outcome measures, the most common of which is a measure of cognitive function, the ADAS-Cog (cognitive portion of the Alzheimer’s Disease Assessment Scale).

The ADAS-Cog has consistently shown a 2.5-3.5 point benefit of CI’s over placebo, a difference that is statistically significant. But is this clinically significant? Since the highest possible score on the ADAS-Cog is 70, a 3 point separation between drug and placebo represents about a 4% improvement in the scale. Because of concerns that such a small improvement in cognition might not be clinically relevant, most of these studies include a global clinical rating of improvement as a primary outcome, generally the “Clinician Interview- Based Impression of Change scale with caregiver input” (CIBIC-Plus). This is a 7 point scale, ranging from 1 (markedly improved from baseline) to 7 (markedly worse). The drug-placebo difference is typically 0.3-0.5, or, in terms of percent, about 6% benefit over placebo. Again, while this is a large enough separation to achieve statistical significance, its clinical significance is debatable. (For an excellent review of these studies, see N Engl J Med 2004; 351:56-67.)

Nonetheless, many of us have heard that CI’s have been shown to delay nursing home placement by a good six months, and, if true, this certainly appears clinically relevant. Unfortunately, it turns out that there is no solid empirical basis for this truism. The “six-month delay” is an estimate based on the fact that in untreated AD, the ADAS-Cog worsens by 7 points per year, and so the typical 3.5 point benefit due to CI’s slows the course of AD by six months, presumably delaying nursing home placement by that amount as well. This logic isn’t bad, but the Lancet study was the first to empirically test whether this assumption is accurate.

Enough methodology! Here are the Lancet results:

1. At the three year endpoint, there was no difference between Aricept and placebo in rate of institutionalization or progression to severe disability.


2. On the MMSE (Mini Mental State Exam), patients on Aricept consistently averaged about 1 point higher than patients on placebo, throughout the study.


3. There were no significant differences between Aricept and placebo in behavioral and psychological symptoms, time spent by caregivers, or formal care costs

This is the short version of the results. The study was complex and some of the interim results reflected better on Aricept. But the bottom line is that the first long-term controlled study of a CI for AD showed it to be only marginally effective for cognition (assuming that a 1 point separation on the MMSE is marginal), and relatively ineffective for delaying institutionalization or disability.

So, what does this data mean for our practices? Should we stop prescribing CI’s for patients, on the theory that the cognitive benefits are extremely slim, and the long-term functional benefits are nil? Well, according to the major government medical institute in Great Britain, that is exactly what we should do. The National Institute for Health and Clinical Excellence (NICE) recently issued draft guidelines on the use of CI’s in dementia (free download at http://www.nice.org.uk/page.aspx? o=104058). Their recommendation was that Great Britain should remove all CI’s and memantine from the NHS formulary because of a lack of demonstrated cost effectiveness.

That’s a pretty radical move, and one that is being bitterly disputed by medical societies throughout Britain. Much of this battle is being fought in a rarefied landscape of pharmacoeconomic statistics, and we will not drag you through that terrain!

At this point, the bottom line is that while CIs do yield a small cognitive benefit in patients with AD, they are expensive and they are unlikely to delay either nursing home placement or the development of severe disability. On the other hand, given the recent study completely discrediting Vitamin E as a memory-enhancer (N Eng J Med, 2005, 352:2379-2388) we have no other medication to offer our demented patients. For this reason alone, CIs are likely here to stay, at least until something better comes along. Of course, you could resist the temptation to whip out the prescription pad and focus your efforts on behavioral recommendations for the patient and family that are less costly and sometimes as effective, such as encouraging daily structure and providing environmental cues. (There are some good specific suggestions in a caregiver handout at http://alzheimer.ucdavis.edu/careg/media/pdf/dementia handout.pdf.)

TCR VERDICT: CIs: Minimal benefit, but what else do we have?