Statistical challenges in the evaluation of surrogate endpoints in randomized trials