I believe that halabi survey of the 02B data is very recent. Gold first described this data, qualitatively, in the CC after the approvable letter. I believe David Miller asked him at the time *when* they did this analysis and Gold remarked it was a very recent endeavour.
I'd wager no less than 350 patients in that analysis... it may even be as reasonably real time as a blinded trial allows.
edit:
Actually, those data have to be practically real time. They're based on patient characteristics at enrollment, and those are disclosed up front.