Reduced sample sizes for atrophy outcomes in Alzheimer's disease trials: baseline adjustment