Thursday, October 20, 2005
Risk of death with atypical antipsychotics in dementia
This week's JAMA has a meta-analysis examining the increased risk of death with atypical antipsychotics in demented patients. The authors looked at only randomized placebo-controlled trials (although a couple of trials had haloperidol arms) for a total of ~3300 patients. Generally, all the meta-analysis i's & t's were dotted and crossed. The OR of death with atypicals vs. placebo was 1.54 & the authors calculated a number-needed-to-harm of 100 (with 95% CI NNH ranging from 53-1000). They point out that the NNT for atypical antipsychotics in dementia is 4-12 & that, then, for every 9-25 patients helped in these trials there was 1 excess death. This, I think, makes the NNH of 100 seem a bit less acceptable. None of the novel antipsychotics were particularly worse than the others. They did briefly analyze the mortality with the haloperidol patients and compared them to placebo & found that those patients receiving haloperidol died more too. It's important to note as well that most of these trials followed patients for less than 12 weeks, meaning that the mortality difference was noticable in this relatively short time.
So is it time to bring out the principle of double-effect for antipsychotics in dementia? That these are drugs that may improve people's quality of life but also may hasten death in a few? There are probably a few geriatricians amongst this blog's readers & I'd be interested in any comments.
On a mundane note I'm travelling this coming week &, depending on how busy I am, may not post much. I've noticed that the last week I've been getting quite a few more hits than before and am creeping up to 50 a day sometimes which is astonishing to me. Thank you, all who read. Please leave me comments if you'd like: in general I'm curious as to who, besides my mother & immediate colleagues, reads this blog, as well as any general feedback, criticism, etc. You can leave comments anonymously if you want. I'm going to try to lengthen the "recent posts" section to make reviewing past posts easier especially for new or infrequent readers. I've also decided that I'm really angry at the Lancet website which makes you log-on just to read a simple abstract (unless you're visiting from a proxied server) & am going to try to put those links into PubMed. The problem is that on occasion PubMed can be days or weeks behind electronic publication so sometimes there's nothing in PubMed to link to.