The need to show minimum clinically important differences in Alzheimer's disease trials.