Small sample sizes in clinical trials: a statistician's perspective