That's a GREAT question! There are many different ways to do randomization. The best way to ensure balanced arms is via stratification during randomization, for example stratify by ECOG, disease stage, race etc to ensure different arms are balanced by these basic factors. It doesn't seem they stratified in their randomized trials, which is why many on this board complained the so called imbalance for example ECOG between arms. Whose fault is that? I guess if they were stratified, no one could find excuses to explain away things anymore.