[Pre-randomisation in study designs: getting past the taboo].