Whole Brain Atrophy and Sample Size Estimate via Iterative Principal Component Analysis for Twelve-month Alzheimer's Disease Trials

Abstract

In this study, we used Iterative Principal Components Analysis (IPCA) to characterize twelve-month wholebrain atrophy rates in 125 Alzheimer’s disease (AD) patients, 288 subjects with amnestic mild cognitive impairment (MCI), and 167 elderly healthy controls (HCs). We compared IPCA to the widely used Brain Boundary Shift Integral (BBSI) and clinical measures of change from the same subjects. Both IPCA and BBSI techniques correlated with changes in Clinical Dementia Rating Scale-Sum of Boxes (CDR-SB), Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) and Mini Mental State Exam (MMSE). Both IPCA and BBSI–based whole brain atrophy differed among AD, MCI and NC groups and had comparable statistical power. For AD, we estimate the need for 70 patients per group using IPCA and 80 using BBSI to detect a 25% atrophy-slowing effect over twelve months with p=0.05 and 80% power, compared to 514, 636 and 843 using CDR-SB, ADAS-Cog and MMSE. For MCI, we estimate the need for 128 patients per group using IPCA and 136 using BBSI compared to 798, 5392 and 3531 using CDR-SB, ADAS-Cog and MMSE. As an alternative for characterizing whole brain atrophy, the fully automatic IPCA procedure offers sample size precision comparable to that of BBSI. Keywords: Alzheimer's disease, iterative principal component analysis (IPCA), sample size, statistical power, structural MRI, whole brain atrophy.