Richard Smith, former editor of the BMJ, discusses the problems that complicate and sometimes undermine scientists' ability to perform or interpret clinical studies, even the highly prized randomized controlled trial.