Perhaps the reformulation was the issue (or at least part of it), but I would tend to look elsewhere.
First of all, becuase the trial had 2 in interiums (at known N) and a known enrollment curve, the blended trial curve is reasonably known. And we also know there will not be much seperation of the 2 curves (it was a large trial, so a decent seperation would have been enough). I would guess the medians will come in at about 26 months control vs 30 months Stim.
This value for Stim survival would be inline with both the P2 subset and reformulation safety study. If so, the drug did perform the same as the orriginal formulation. In that case, the "issue" was that the conntrol arm lived longer than in the P2 subset. There is little historical data to go by.
If we look back at the P2, one thing of note was the the healthier Stg3LR group in the control arm did not live longer than sicker Stg3 nonLR and Stg4 patients. One has to wonder, did the subset just detect random chance here?
Hopefuly Merk KGA will present the results at ASCO.