But comparing the two ISIS-FXIRx arms to each other is a different experiment with a different false-positive rate.
Ahh, the subjectivity of statistics. I don't think even Fleming would suggest ignoring a p value that low just because it wasn't a primary endpoint. Exactly how much credence to give it is, of course, subjective. (In a more extreme version - Would you ignore a p value of 0.00000001 just because all the alpha was already spent? Again, I suspect even Fleming would approve such a result as proof of efficacy if it were apriori reasonable.)
Market Data 
Markets 
AI
