You only want to enroll patients for the purpose of completing the trial. If you succeed in completing the trial with fewer than prepared-for patients, that's a good thing. 10,000 other patients can get access to the life-saving treatment in the approval time saved by not holding out for 100 additional trial patients enrolling when you didn't need them. Everyone else can get compassionate use between then and actual approval (and reimbursement/coverage scheduling), remaining control patients get to cross over (as continuing to withhold treatment would be unethical), etc.
As for early halt during accrual versus afterward, it does happen, there's a long history of it. Not only has it happened, but it's even happened with fewer than 1/3rd of the planned patient population enrolled (e.g. the PTCA trial in '95 which stopped with 1500 out of a planned 4800 patients).
Furthermore, this trial is in a particularly unique situation favorable to early halt relative to accrual completion because of how new enrollment (but not ongoing treatment) was paused from '08-'11. Patient count may be low but observed patient-months and events are disproportionately large to that count (if it had just started in 2011 instead of 2006). Even if the 33 existing patients had far longer PFS than SoC, their first treatment schedule (the 120 weeks) had already ended by the time enrollment resumed and was years past by the time the first interim was triggered. My point?
The first possibility is that they at long last, sad but likely, contributed to the 66 events. So it didn't really matter how slow enrollment has been because it truly started so long ago.
The second possibility is that they've been progression free and alive for so long that the distance between the K-M curves is enormous and however else the 66 events were comprised (e.g. resumed enrollment was slightly further along than we expected and SoC performed...well, normally for SoC), the treatment advantage has already proven huge.
Either way, there have been a great many patient-months observed in the study.
Obviously this is a very bullish scenario. It's worth repeating that by far, historically, the most common decision is to simply continue. I'm just saying that this trial's particular, elongated history and careful design makes it somewhat (greater than 0%) more likely than any given other trial to achieve this unusual early outcome (if the treatment works).
