Correct, the “bear” theory is the median went down from the 331 pool (2018) to the 232 pool (2022) because, in their words, the 232 did worse than the 99.
This is likely incorrect, imo. The reason the 232 pool went down by 0.7 months is because the five or six “dropouts” that were still being treated as censors were no longer part of the 232 group pool, because they were in the 99 pool, after breaking the 331 into its constituent parts. So the 232 group got reduced early, and this gave “bears” overconfidence.
On the other hand, the only way to get those censors removed from the group of 99 is to replace them with the actual event dates. As I’ve stated earlier, both the km curves and therapeutic reasoning point towards these censors being earlier eventors — Aka: before their median. As I’ve also said, Pub Med articles state the only real way to untangle ltfu censors is to find the true data points. And as I’ve said, I can’t imagine this can’t be done here, but until recently, due to trial maturity, QALY valuations and parallel ECA groups, there’s been no reason to locate these data events.
Now there is, and I won’t bore you with the details, because they involve further speculation, but my guess is the 99 curve falls closer to the ECA curve, but above it (crossover impact), and the treatment arm could be stat. sig. there as well if all data points were to be located. I think they will be located.
Just my processing.
News 
Market Data 
Discover 
