About a month ago my doctor told me my diabetes medicine wasn’t working as well as it used to, so he added Avandia into the mix. It worked very well – after a week or so my blood sugar had dropped to nearly normal levels.

A week after that the media started reporting that this stuff could make your heart attack you. According to a study the risk of myocardial infarctions went up 43% for Avandia users.

I wasn’t too worried, because a 43% increase isn’t much when it comes to epidemiology. Statistics are a yardstick, not a micrometer. Even studies that are well done and honest don’t yield exact numbers, just approximate increases in probability. Because of that, and because there are errors in every study, increases of less than 100% are usually not worth getting excited about. But still, I was taking the stuff, so I set out to see if there was any validity to the claims.

The first article I found said it was based on combined information from 42 studies. That means it’s a meta analysis. That means it’s probably crap. Meta analysis, a study of studies, is the most difficult kind of study to do correctly, and the easiest to fake and manipulate. They are almost always garbage.

Other articles explained that Avandia increases water retention, and that can, in turn, cause a slight increase in heart attack risk. OK, that sounds plausible , but what were the real numbers?

I found the entire study on line, and discovered it was even more useless than I expected.

The Risk Ratio of increased heart attacks (RR) was 1.43, which translates to a 43% increase. But the Confidence Interval (CI) was 1.03 to 1.98. Although the CI is technically not the same as a margin of error, it works the same way. The real number could be a .03% increase. Or a 98% increase. Or any number in between.

If the lower bound of the CI had been 1.00 instead of 1.03, the RR would have been statistically insignificant. The study would probably have never been published, and we wouldn’t be reading all these scary articles.

But wait, there’s more. In the methodology section of the report we’re told: “Six of the 48 trials did not report any myocardial infarctions or deaths from cardiovascular causes and therefore were not included in the analysis because the effect measure could not be calculated.” A half dozen studies showed no heart attacks so they left them out. This is the very definition of cherry picking.

And then my mother-in-law, who is also a diabetic, sent me this article. It fills in even more details and shows just how appalling this study was. It explains that there were 14 additional deaths out of 25,000 people taking Avandia. Statistically, a variation that tiny means nothing.

Oh, and the Doctor behind it? He’s funded by Takeda. Takeda manufactures Actos, a drug that competes with Avandia.

I think I’ll keep taking the stuff.

If you’d like to become an expert at ferreting out dishonest medical studies, read Statistics 101 and Statistics 102 on The Facts.