To the Editor: Drs Hennekens and DeMets1 argued that large RCTs designed to test a priori hypotheses are the best strategy to detect small to moderate benefit or harm from drugs. They suggested that the results of meta-analyses or subgroup analyses of RCTs should be considered as hypothesis-generating and formally tested in suitably powered RCTs. This is a sensible strategy for evaluating potential benefits of a drug but may be impractical for exploring potential harms.
Bolland MJ, Grey A, Reid IR. Evidence From Randomized Controlled Trials, Meta-analyses, and Subgroup Analyses. JAMA. 2010;303(13):1253-1255. doi:10.1001/jama.2010.366