All had relatively small sample size and this is the only advantage I see in these protocols when used for screening drugs.
Yeah, if you believe that most drugs are worthless and you want to screen a large number to determine which few, if any, MIGHT have some powerful clinical benefit then this methodology is pretty efficient. As you note, when filtering for a powerful drug it uses a relatively small number of patients. However the savings is a lot less if you are filtering more broadly - i.e. not just looking for the next Crizotinib. E.g. in the trial you pointed to they set the clinical utility boundary as 30% below the placebo treatment and, at least in cancer, there are many successful drugs with less than 30% efficacy. If you move to a more typical 20% the trials would get substantially bigger.
PS Note that I would make sure that they were placebo controlled - no reason that they cannot be. Just wasn't done on the trial you happened to point to.
News 
Market Data 
Discover 
