SAPs
The concept of what is and what isn't prespecified in a statistical analysis plan is somewhat murky. Most companies don't reveal their statistical analysis plans, but Genta's SAP for the melanoma trial did get released and it was informative.
Bottom line is that SAPs are an exhaustive laundry list of various analyses. Cox models can be prespecified, and we know for certain that many subgroup analyses are also prespecified. However, we also know that companies have an utterly abysmal chance of receiving approval based on subgroup analyses that were prespecified in the SAP. There are multiple reasons for this, but the fate of such applications are well known.
So the fact that an analysis was "prespecified" is not overly informative insofar as its suitability to be the main efficacy driver in an NDA / BLA application.
I think the best rule of thumb is to assume that the important prespecified analyses are the ones that the sponsor discloses while the trial is running. Companies readily disclose that their currently running / enrolling phase III trial has a primary endpoint of overall survival or PFS or response rate or whatever. That is probably when companies are most candid, and it is best to rely on those endpoints as the prespecified ones that will fly with the FDA.
If you hear of a certain type of analysis for the first time on the day of data release or a month or two after data release, then you know that the "prespecified" analysis that is now being touted was one of dozens that are written into SAPs in order to try and explore a treatment effect, if any.
Market Data 
Markets 
AI
