OBJECTIVE: To identify cognitively normal individuals at risk of Alzheimer disease (AD) based on cerebrospinal fluid (CSF) Aβ1-42, and to determine rates of cerebral atrophy. METHODS: Control subjects from the Alzheimer's Disease Neuroimaging Initiative with CSF and serial magnetic resonance imaging (MRI) were dichotomized on CSF Aβ1-42 (normal control [NC]-high > 192 pg/ml; NC-low ≤ 192 pg/ml). Baseline and 1-year MRIs were registered, and brain, hippocampal, and ventricular volumes and annualized volume changes were calculated. Baseline characteristics, CSF profiles, neuropsychology, brain volumes and atrophy rates, and APOE, PICALM, CLU, and TOMM40 genotypes were compared. Sample sizes to power presymptomatic clinical trials based on rate of atrophy were calculated. RESULTS: Forty of 105 (38%) were classified as NC-low, and 65 (62%) as NC-high. There were no differences in age (76.3 ± 5.1 vs 74.9 ± 5.1 years), gender, brain volumes, and all but 1 cognitive score (Trails B; p = 0.015). The NC-low group had higher tau (p = 0.005) and p-tau (p < 0.001), and was more likely to be APOE4 positive (48% vs 11%, p < 0.001). The NC-low group had significantly higher whole brain loss (9.3 vs 4.4 ml/yr, p < 0.001), ventricular expansion (2.04 vs 0.95 ml/yr, p = 0.002), and hippocampal atrophy rate (0.07 vs 0.03 ml/yr, p = 0.029). Baseline Aβ1-42 level was strongly correlated with rate of brain atrophy only in the NC-low group (p < 0.001). Using 141 (95% confidence interval, 86-287) patients per arm provides 80% power in a 1-year treatment trial to show 25% slowing of brain atrophy in the NC-low group. INTERPRETATION: A significant percentage of healthy older adults have CSF profiles consistent with AD and increased rates of brain atrophy, suggesting that they may be in the earliest stages of neurodegeneration. Brain atrophy may be a feasible outcome measure for AD prevention studies.