were considered acceptable if they were applied equally to both intervention arms. Trials with more than 1,000 subjects were selected to minimize publication bias, because pub- lication of such large studies is unlikely to depend on the magnirude and direction of their results (27)