Agreed that your explanation looks good but the math estimations are within range of a stealth crossover. For example, 232 treatment+54 regular cross+12 stealth=298=90.03%. I think there were 12 censors or lost to followup counted at one point if not mistaken and the 232+54 gives you the 86% mark reported. The numbers are really tight so your explanation seems more reasonable but... Best wishes.
Thanks Lykiri, unfortunately by December 2016 we already had the 331 patients with 232 treatment and 99 placebos set in stone. Dr. Liau knew that the randomization of 2.3:1 was permanent. If there were already 66 placebo crossovers in December 2016, we would never have an additional 46 noncrossover placebos because there were never as many as 112 placebos. Had that been possible than yes, we could have had 14% placebos. But then with only 331 patients and let's say 111-112 placebos, we would have ended up with only 219 or 220 trial patients instead of the 232 we ended up with.