I think we're already in a realm of numerous unknown/speculated variables so any exact number from me is likely meaningless. However, if I had to say something, I'd say that if it doesn't happen at the first interim, it's twice as likely to occur at the second interim.
Whether this particular DMC gives a damn about collecting sufficient data for the secondary endpoint (OS) to reach significance is perhaps the greatest wildcard - I've seen it go both ways. I think with adequate PFS advantage as a predictor of eventual OS advantage, most of the members won't care, but I could be wrong about that.
I do think it is quite unlikely that the sample size will be increased at the stat-only interim if we get that far. Either it's going to work well enough that the PFS advantage will already be plenty significant (just look at the thick cushioning already built into the trial parameters), or it's going to work poorly enough that anything short of doubling the trial size would be potentially futile and the DMC would have a hard time justifying the trial's extension (and thus keeping more patients out of other ongoing trials, potentially wasting company money, etc.)
I like that they gave themselves the option of another increase just in case, but I honestly think they did so knowing it would probably never be exercised (not only because of the sizable statistical cushion at 110 events [p-value could be double the plan and still significant] but also due to the likelihood of a halt at the 1st or 2nd interim [since earlier results had thrice the advantage assumed for this trial])
Market Data 
Markets 
AI
