The correlations are not post hoc, they are PM AI analysis of extended trial patients.
What you seem to fail to understand is that post hoc analysis applies to ANY data where the endpoints are not pre-specified. What happens in that case?....they do data dredging through thousands of possibilities pretty much ENSURING positive results. AVXL loves to do multiple testing and then treat it as a single test. Just look at the p-values they assigned to their POST HOC concentration-effect data fiasco. Some will only start to realize how abhorrent that analysis is AFTER the hammer falls.
This dredging feature applies no matter how hallowed the method (i.e. AI).
Market Data 
Markets 
AI
