O3‐14‐05: How could potential selection bias impact the analysis of correlates of cerebrospinal fluid biomarkers? the memento cohort

subjects. Methods: We included 202 non-demented (CDR 0.5), amyloid-positive (CSF amyloid-b1-42< 640pg/ml) patients from the Amsterdam Dementia Cohort with available clinical followup ( 1 year) and baseline 21-channel resting-state EEG. Fast Fourier Transformation was performed on five 8.2-sec epochs per subject to calculate parieto-occipital peak frequency and global relative power of each frequency band (delta 0.5–4Hz, theta 4–8Hz, alpha 8–13Hz, beta 13–30Hz). Outcome measure was clinical progression, defined as CDR change 0.5. Cox proportional hazard analyses were performed for each box-cox and z-transformed spectral parameter with age and gender as covariates. Results: Mean age of the cohort was 67.767.8SD and 49% was female. 105(52%) patients showed clinical progression after a median follow-up of 2.0 years (range 0.4– 6.3). Follow-up for stable subjects was 2.1(1.0– 8.1) years. Median (range) relative power was 0.30(0.59) vs 0.30(0.47) in the delta band, 0.12(0.30) vs 0.12(0.42) in the theta band, 0.27(0.51) vs 0.28(0.63) in the alpha band, and 0.19(0.36) vs 0.18(0.40) in the beta band for subjects who clinically progressed vs stable subjects. Median (range) peak frequency was 9.5(6.7) for subjects who clinically progressed and 9.5(5.3) who remained stable. In the total sample, EEG spectral parameters were not related to clinical progression. When we stratified for CDR, higher relative theta power was related to clinical progression in patients with CDR 0 (n1⁄4 60; HR[CI95%]1⁄4 1.4 [1.0 – 2.2]), but not in patients with CDR 0.5 (HR[CI95%]1⁄4 1.0 [0.9 – 1.3). This effect seemed largely attributable to the younger individuals (CDR 0, age 65; n1⁄4 24; HR[CI95%]1⁄4 6.4[1.4– 29.2]; compared to CDR 0, age>65; HR[CI95%]1⁄4 1.2[0.7 – 2.1]). Conclusions: Higher EEG-derived relative theta power was related to clinical progression in amyloid positive, cognitively normal individuals younger than 65 years old. In future studies, we will focus on connectivity based analysis of the EEG signal.