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- location of such large studies is unlikely to depend on the magnitude and direction of their results (27)