Distraction from randomization in diagnostic research.