The effects of excluding patients from the analysis in randomised controlled trials: meta-epidemiological study | Synapse