On the Use of Surrogate End Points in Randomized Trials