What concerns me is that they hadn't check to see if the doctors...

What concerns me is that they hadn't check to see if the doctors can agree on a diagnosis from the beginning. Even the failed trial from last year should have picked up this issue. A 67% agreement rate from 2 doctors is getting pretty close to just pulling a diagnosis out of a hat (50%). Under these conditions even a perfect test would fail because the two doctors would disagree with a perfect test 33% of the time.

This diagnosis problem is not solved by bringing in a third doctor since it doesn't detected when the two doctors both agree and are both wrong. Points to a very poor study design.