And yet, how many times do we see a randomized trial fail because patients on the control arm did better than expected? Whether that is because the care in clinical trials is better than that generally available, or because the patients in the trial have different characteristics than those for which your "pretty much know results" obtains, or for some other reason, I don't know.
I admit, all of this is theory, certainly if I or a family member needed to go onto a trial, I would not be happy about the chance of getting a placebo. I'm sure as you know, in some trials --e.g., Dendreon -- the trial allows cross-over to the treatment arm after progression, some consolation. But that makes it harder to prove the treatment works, since it is, to some extent, competing against itself.
This is all theory since the FDA requires at least one randomized phase III trial, so its not as if companies have a choice.
I think this all started with the observation that the results so far were not based on a randomized trial, and so were less reliable. I am pretty sure David Miller would agree with the first part of your last sentence, that there should be randomized trials early on.
