Nh, the problem is that there is no such thing as a historical norm to compare against.
Consider SUNRISE, do you know for certain that the control arm did not perform better than listed previous trials because of new agents available? Not just the PD-1's but also various target inhibitors that have become available since the historical data you compare to?
And if that is what made the control arm perform better, did it have an effect on the Bavi arm? And if so, do you really know anything about how Bavi itself performed?
And an even bigger issue is inclusion criterion and population selection. If all a trial has to do is beat previous trials, why not incrementally tweek the population to be healthier?
I do not disagree with the point that controlled trials as run now have issues. But to date, nobody has figured out how to do it any better.
CP: On your second point quoted by nh, can I assume you meant BTD (not BLA)? If so, I think the issue is probably moot.
Market Data 
Markets 
AI
