How well can you predict the outcome of clinical trials? Not as well as you may think