August 11, 2007

New Meta-Study on Avandia

Posted by Eric at 12:22 pm | Category: Medicine, News, Pharmaceuticals

In the latest entry into the whole Avandia (a.k.a. rosiglitazone) deal, yet another meta-analysis published in the Annals of Internal Medicine shows no statistically significant increase in cardiovascular risk from Avandia, further giving weight to the quote from Ernest Rutherford, “if your experiment needs statistics, you ought to have done a better experiment.” Or weight to any number of other quotes, including Mark Twain’s, “there are three kinds of lies: lies, damned lies, and statistics.”

So, in conclusion, drugs have side effects, sometimes which are risky. (oooh.) Maybe Avandia has some, too, but no one knows. All Steve Nissen’s study shows is that we should conduct a clinical trial specifically to address this issue before making any more definitive conclusions.

Follow the comments thread by the RSS 2.0 feed. You can leave a response, or trackback from your own site.

3 Responses to “New Meta-Study on Avandia”

  1. Apollo Says:
    August 11th, 2007 at 10:06 pm

    While in principle I agree, conducting a new, sufficiently-powered clinical trial specifically addressing cardiovascular risk (MI or cardiovascular death) on rosiglitazone could take another decade. At least according to the FDA, the RECORD trial is too underpowered and not designed in a manner that can offer any more meaningful information on cardiovascular risk than it is doing now.

    NIDDK scientists did a meta-analysis very similar to the one you posted, and I went to a meeting where they discussed it: with various continuity adjustments, the p-value would bounce just within or outside statistical significance. The FDA presented its own meta-analysis at the advisory committee meeting I attended, and they concluded that rosiglitazone has an increased risk of myocardial infarctions over placebo (non-use), but not over active comparators (metformin and a sulfonylurea). However, they were unable to conclude whether or not rosiglitazone had an increased risk of MIs over the active comparators as the confidence interval was too wide.

    No one paid any attention to GSK’s new meta-analysis presented at that meeting which claimed that rosiglitazone and pioglitazone had the same cardiovascular safety/risk profile. It was eerie: it was almost as if that part of the meeting never happened; people just seemed to ignore it.

  2. Eric Says:
    August 11th, 2007 at 10:20 pm

    I guess the problem with GSK was that their scientists presented badly (as you said) and that people presume that drug companies and anyone connected to them are just lying and outright biased, which is unfortunate. After all, drug companies do publish trials all the time that prove their own drugs don’t work, since most drug candidates fail based on the company’s research.

    I guess if they really wanted to, the FDA could suspend sales of rosiglitazone and put the condition of a new clinical trial before they sell it again, but I guess they decided it wasn’t worth the tradeoff of barring people from access to the medicine.

    I really don’t trust meta-analyses. I mean, clinical trials are complicated enough beasts to analyze; pooling multiple, differently designed clinical trials, and many small measurements, at that, to make a meta-analysis as one big measurement just seems like bad statistics to me.

  3. Apollo Says:
    August 11th, 2007 at 10:35 pm

    Yeah, meta-analyses are weak and not convincingly conclusive. I think that’s the main reason why the FDA didn’t pull the drug: they didn’t want to set a precedent of using just a meta-analysis to determine drug safety, although there was considerably more data presented to the FDA than just meta-analyses (though not in the eyes of the public or popular media) with respect to their black-box warning decisions.

Leave a Reply