As if one were needed, here's an example of how rough the state of current oncology therapy is today. Avastin, the antibody-based therapy from Genentech/Roche, had been approved (conditionally) for advanced breast cancer, based on a study showing about a five-month benefit in tumor growth. (Everyone should already know that such numbers, for many types of cancer, are indeed enough to get an indication approved, and everyone has, I'm sure, already decided what they think about that.)
But the approval came with a requirement to follow up on those results. For one thing, the study that led to conditional approval didn't show much of a survival benefit, making the approval itself controversial at the time.. The follow-up work has shown that those initial results were right on target. For metastatic breast cancer, Avastin has something like a month-and-a-half survival benefit. That probably doesn't outweigh the risks, and the FDA is seriously thinking about revoking that earlier approval.
Based on these numbers, I think that they should go ahead and do that. The whole point of conditional or accelerated approval is that it can go either way when the harder numbers come in, and in this case, it seems pretty clear that the benefit isn't there. No one cares about tumor growth if it doesn't affect survival or (at the very least) quality of life. And in this case, the later studies have suggested that even the earlier tumor growth numbers were too optimistic. You have to be willing to abide by the evidence.
Because of Avastin's high cost, this is probably going to turn into a rationing-health-care argument - in fact, it probably has already. But I'm not even talking cost here. Avastin, by the evidence we have, does not seem to help advanced breast cancer patients. It wouldn't help them even if it were free.