楊宗翰 說 Jun 06, 2010 10:37AM Ideal: adequately powered, appropriate comparators, generalisable, cluster randomized, hard patient-based endpoints