"On the flip side, control patients were older and had a lower KPS on avg"
That's not the flip side: both measures as you present them also favors the rindo arm. Of the three measurables you mentioned- extent of resection, age, and health (KPS score) all would favor patients living longer. Even if rindo had no effect on their outcomes, one would expect patients in the rindo+Avastin arm to outperform patients in the Avastin arm.
Now one can make guesses about how much of an effect these imbalances had and attempt to compensate for them, and the data still look promising.
Of course there are still many important factors that were measured that also have a big effect on survival. With such a small population, it is unlikely the arms were balanced in those factors also.
Taken together, its a nice pilot study that hints at efficacy but not enough for FDA approval.
In this case franks profit report is probably accurate. Careful followers of the Frazzano Phenomena will remember he changed his style of accounting for this position.
Instead of lumping all profits in a stock since birth into the cost basis of his current shares, the huge drop in his original EXEL position after a failure in a prostate cancer trial forced him to switch to a more traditional FIFO accounting for shares at low cost. That nasty hit from a few years ago disappears leaving only the recent gains.
Everytime you test for efficacy with an interim analysis you make the closing test more strict. It basically compensates for the extra chances you have to get lucky and meet the required p value even if the drug was not effective.
The final p value depends on how strict you make the required p values to stop at the interim checks.
Here they made them fairly strict so that the final required p value does not take a big hit.
The final N does not affect the p value you seek. A bigger N does make it more meaningful, however.
See, for example, the REACT study with small N and multiple peeks at the data and the FDA response.