Good statistics, bad statistics part II

I wasn't going to use this example for my second installment on statistics, but when I read it on Sandy Szwarc's site I thought it was perfect for presenting one of the big issues with so many studies we hear about in the media. (Ms. Szwarc, I hope you do not mind at my continued references to your work.)

The purpose of the base study referenced in the article? To investigate if a certain drug can alleviate weight gain caused by taking an anti-psychotic. (Trust me, it's there; just keep reading.) The money quote from Ms. Szwarc regarding that study is this:

So, they tested 3 male schizophrenic patients (average age of 22 years) hospitalized for acute psychotic episodes by giving betahistine along with their anti-psychotic medication (olanzapine) for 6 weeks.

Three guys! Three guys comprised the study! Nothing statistically significant could be gleaned from such a small study and such a small study should never be considered a basis for any decision except that further study is required. I don't care if the researchers claim it was statistically significant; I simply don't buy it. That sample is way, way too small, even for a cell or a cluster.

And yet, as Ms. Szwarc demonstrated aptly in her article, the pharmaceutical firms are going crazy, wanting to work with the listed drug and media articles going ga-ga over this super fantastic drug (to borrow a phrase from Manolo).

I really worry that soon the media will be going crazy because a study came out based on two persons or even just one person that demonstrate the next Practically Perfect Pill™. You'll never hear about the study; just how great it could be, should be, how we should be behind it now.

Sad, sad, sad.