Email Post: Generalizing the per-protocol treatment effect: The case of ACTG A5095