CLSN, ONTY, AVEO -- iwfal
It's interesting you list these three because they're the three I've spent the most time modeling.
It's hard to pound the table on CLSN because of the lack of comparable trials. I'm bullish on that one just because of the sheer magnitude of the implied median PFS, but have to concede it's possible that control median PFS is way higher than expected.
With ONTY, I don't agree that the trials that were "more comparable" have the longer medians. In fact, the trials with the longer medians are mostly single-arm Phase II trials. The exception is that SWOG trial that had two cullings to get down to the superhero population that didn't like gefitinib, if memory serves. The Bernstein report (and the Seeking Alpha thing) debunked that big MST pretty well, I thought. There are some bigger Phase III trials to look at, and in none of the more plain vanilla big trials do we see the kind of control MST we'd need to shoot down the START trial.
I suspect in the ONTY case, however, that the real issue is that many have mismodeled the first interim, and have drawn the wrong conclusions from it. We'll know soon enough (supposedly Q1 2012).
On AVEO, we've had the back and forth. I conceded that your example could be accurate, but that "the odds" favored success based on the calendar. (If I need a much higher LTFU rate than the Phase II trial, and a much higher median PFS for Nexevar than we've seen in other trials, then Lady Luck leans the right way.) I think we're supposed to hear something in Q1 2012.
At any rate, I think these are three good tests of the method. I look forward to assessing the outcomes as a gauge of whether "the numbers game" is actually useful.
Regards,
TGW
Market Data 
Markets 
AI
