The Authors Respond

    loading  Checking for direct PDF access through Ovid


We agree with Shinozaki et al. regarding the expected bias of an intention-to-treat analysis to estimate the per-protocol effect (a problem well understood by trialists) and the possibility of lack of generalizability due to nonrandomized concomitant interventions. We prefer to reserve the term “bias” for lack of internal validity.
    loading  Loading Related Articles