Biomarkers for predicting cognitive decline in those with normal cognition.

Most studies evaluating Alzheimer's disease (AD) biomarkers longitudinally have studied patients with mild cognitive impairment (MCI) who progress to AD; data on normal subjects are scarce. We studied which biomarkers best predict cognitive decline on the Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) among those with normal cognition at baseline, and derived cut points to predict decline. We studied 191 subjects in the Alzheimer's Disease Neuroimaging Initiative (ADNI) who had normal cognition at baseline, 2 + visits (mean follow-up 3.1 years), and data on neuropsychological tests, cerebrospinal fluid (CSF) biomarkers, and structural MRI. We used repeated measures linear regression of log ADAS-Cog on age, race, gender, education, APOE4 status, baseline biomarker values, and follow-up time; an interaction between biomarker and time assessed predictive power. Neuropsychological tests did not significantly predict ADAS-Cog decline, while both MRI variables and CSF biomarkers did; CSF markers were the strongest predictors. Optimal cut points for baseline CSF markers to distinguish decliners were < 220 pg/ml (Aβ42), ≥61 pg/ml (t-tau), ≥21 pg/ml (p-tau), ≥0.31 (t-tau/Aβ42), and ≥0.10 (p-tau/Aβ42). For progression to MCI/AD (n = 28), the best markers were t-tau, t-tau/Aβ42, and p-tau/Aβ42, with optimal cut points of 58, 0.31, and 0.08, respectively. The optimal cut points across all markers and cut points predicted decline in ADAS-Cog, as well as transition to MCI, with a 65% accuracy. Our findings support current models of AD progression and suggest it is feasible to establish biomarker criteria to predict cognitive decline in individuals with normal cognition. Larger studies will be needed to more accurately characterize optimal cut points.