There's an old saying,"There's lies, damned lies, and statistics." I think the varying reactions to the Avandia study by Glaxo show this to be true.
So who's right?
For my money- I'd like to see more studies present results with "Number Needed to Treat" and "Number Needed to Harm" data. For instance, "This study shows if you treat 50 patients with drug x- you will prevent one death and two heart attacks. You need to treat 1500 people to get liver function problems that are reversible. You need to treat 20,000 people to kill someone with drug x." Instead, I get a lot of statistical static and an editorial that emphatically states I need to prescribe more drug x.