Monday, May 30, 2011

Does Aricept work? yes/no/maybe so

So this is my favorite picture from any study I've ever read on the treatment of dementia. This study is from Mass General's NIH funded Dementia/Memory center. Published in 2008 in the Alzheimer's Disease Associations journal, it is a publication of their longitudinal data. In academic terms, this is a retrospective historical cohort control study looking at three groups: Placebo, Aricept only and Aricept plus Namenda. (cohort=group of people) All prospective, placebo controlled trials that are of decent quality are less than 2 years in length, and all but about 4-5 are a year or less in length. So for this chronic disease, we have very little data about the benefits/harms of using the 4 FDA approved dementia meds long term. This study attempts to answer this question using the best data available.

So on to this study. This picture is from Mass General's decades long database of patient info. The top picture is the placebo group, the middle one is Aricept only and the last one is Aricept plus Namenda. What is being measured is cognition (Blessed Dementia Scale) where the higher the number the greater the impairment. The dark line going across each graph is the line of best fit or approximately the average of each group. The flatter the line, the more preserved the cognitive ability. Each little line with dots is an individual person.

The punch line from the study is that this is the longest cohort controlled data published to date (2008) and not only does it confirm a benefit to medications, but an increasing benefit over time. Let's look at this....

So does Aricept and Namenda work?

1. Yes: In this study you can see that the dark line in the bottom picture (Aricept+Namenda) is flatter than the dark line in the top picture (Placebo). This is consistent with almost every placebo controlled study. By the time you get to the end of the study, the group on treatment has an overall/average score that is better than the placebo group.

2. No (You'd never know part 1): If every patient achieved that average score for the population, you would likely never know the difference. If one person worsened by 6 points (out of 36) on placebo but another person worsened by 4 points, you'd be hard pressed to figure out who kept two of their points. While these results are statistically significant, clinically, very difficult to tell. This picture is also seen when looking at function (Activities of daily living) or any other measure of Alzheimer's. Average benefit is incredibly difficult to tell without looking through the magnifying glass of a researcher.

2. No (You'd never know part 2): But every patient does not achieve the average benefit. That should be clear from the chaos of the light lines connected by dots. There is an incredible variation of the course of dementia whether on treatment or off treatment. Patients improve, stabilize and decline whether they are on treatment or not. Trying to compare how a patient is doing compared on treatment to guessing how they would have done off off treatment to measure effectiveness for an individual patient is futile because you just never know. Regardless of group, you can see in this picture that for any specific individual patient, you can see a matching patient in another group. In other words, a stabilization of MMSE, a decline of only 1 point per year, an improvement of 1 point per year, an improvement in ADL's can all come out of any group.

3. Maybe so: So does it work? The answer is maybe. But you'd never know. It is clear these drugs have an effect in populations, that the response is incredibly variable. They may really work. For a very small proportion, they may work really really well and those improvements may be masked in groups of people. But it should be clear that the average person benefits little or not at all. So why do most people feel that these drugs do nothing? An incredibly variable response which is very very small leaves a lot of us thinking that these medications don't seem to be doing anything most of the time. It's just so hard to tell.

Would I take it if I started with mild dementia? Yes, if I were healthy, were on minimal medications, did not have to sacrifice anything to afford it and most importantly had zero side effects from the meds. Also, I'd want to have a decent quality of life (personally meaning that I could communicate and not be in a lot of pain)

Second conclusion: Be very wary of any treatment of dementia that says it works and publishes their average data between two groups of people. The skeptic in me would want to ask: 1. how would I know if it is working so I can stop it if it is not working? 2. What is the exact benefit and what proportion of people achieve that benefit.


Charlotte said...

What about galantamine? My husband is on that because that is what the VA uses. There are very few studies/comments about whether it is affective. From little things his neurologist says I get the feeling she has little belief in that any of the current drugs really work. She specializes in dementia and works for a clinic that does many of the trials.

joshuy said...

I have not found Galantamine to be much different than aricept etc. Here is where you can find the evidence reviews...

The definitive guideline by the ACP/AAFP can be found here ( with a summary here ( and the very detailed evidence summary here (
NIH's most recent summary can be found here ( It tends to be a little pessimistic.

Ann said...
This comment has been removed by a blog administrator.