The company specifically stated that stroke was not part of the SPA. You are right that the data are not convincing. I would think it raises a red flag, and the FDA would have wanted some non inferiority in the result at least.
Do you have any explanation why stroke is not part of the SPA at all? The sample size in all arms will be 5 times bigger, so a somewhat more meaningful data should result.
Just FYI there is a lot of slop in the words "part of the definition of MI". For example, what if the definition in original trial was exclusively CK-MB>100 but in the pivotal trial it is CK-MB>100 OR changed ST. That could substantially dilute the CK-MB effect if ST and CK-MB are not highly(!) correlated. (Note that I agree that "physician diagnosed MI" benefit counterbalances this to some degree.) The interesting question is exactly what the difference is between the ph ii definition and the ph iii definition. Lots of trials fail on such 'trivial' changes.
For instance the Gusto IIb trial used as the criteria 2 of 3 of the following: CKMB, 2 new Q waves or new regional wall motion abnormalities.
Market Data 
Markets 
AI
