One other possibility is that the placebo MACE rate and the drug MACE rate turned out to be lower than the trial design.  

I clearly remember from talking to DM many months back, that RVX was surprised at how slowly MACE were taking place. My jottings from that discussion are "event rates for the trial were almost half the expected design rate, and the rate really slowed down in the summer of 2018".  At that time, I did not know what to make of it...but in hindsight, it adds up.  It can't be only Apabetalone working well...it also has to mean the placebo MACE rate turned out to be lower than expected.  You can't have only half of the patient population slowing down considerably...it has to be all the groups slowing down.

I have no idea why the realized placebo rate would be lower than trial design.  But from a purely mathematical point of view, it suggests a) MACE rate for the placebo has to be lower than what the trial was designed for, b) that makes it even more unlikely that the Lipitor/Crestor bifurcation can be statistically different.