Two comments on JNJ’s Zytiga trial you posted about:
1. The interim OS analysis was evidently dispositive in JNJ’s opinion insofar as the PR on the interim analysis (from Oct 2010) stated categorically that JNJ intended to submit marketing applications for Zytiga in all major jurisdictions.
2. The FDA may have been more willing than usual to bar crossovers within the confines of the phase-3 trial because JNJ was on the verge of opening an expanded access program for Zytiga.
I'll add some dates to your timeline from the FDA medical review document. 8/20/10 DMC recommends study unblinding - "The protocol pre-specified interim analysis showed an overall survival benefit with abiraterone. The IDMC recommended unblinding of the study and crossing over of patients initially assigned on placebo. FDA concurred with the proposals." 8/26/10 Amendment 3 to SPA to allow unblinding and allow placebo to crossover 9/09/10 JNJ issues PR on study unblinding and allowing placebo crossover 9/20/10 Data cut off for final analysis
That the FDA requested updated survival data doesn't seem so troubling to me given that trigger occurred at 67% info fraction and median follow up was only 12.8 months at the interim cutoff. I think the question here is what took the DMC so long to analyze and make the recommendation from the Jan interim trigger. My suspicion is that the "I"DMC was not so independent from JNJ. Delaying the time to crossover also gave JNJ valuable data on when Zytiga's effect would dissipate - and the p value decimal place did move a couple of places to the right despite the higher number of events. Hence, imo JNJ's decision to allow crossover of 302 after the interim despite OS not being stat sig (which apparently the FDA does not have a problem allowing crossover and did not have a problem with in 301 upon notification).
p.s. I also will note that 301 used standard OBF with standard boundaries ;-)
By my reading the interim analysis did not occur in Jan 2010, but only in August 2010. I don't know why the delay between the cut-off date and the interim analysis date, but it's not something you can blame the FDA for. Given the long lag, an updated OS analysis before crossover seems quite reasonable - no reason to throw away potentially very informative data.