One of the first text I read in graduate school was "How to Lie with
Statistics". Let's say there was 500 people were in the Celebrex sample and 500
persons not taking Celebrex. 2.5% of 500 is is around 12.5. If 1.25% of  control had
heart attacks, you have about 6 people, but 12.5 is 100% more than 6.
Therefore,  you have 100% more occurences in the Celebrex sample. But the real
difference is about 6 people!

Bobby
2 1/2% at
risk is not that alarming to me. Of course, I do not take either Vioxx or
Celebrex.