How does scaling up randomised controlled trials make them more effective?